KSHV infection drives poorly cytotoxic CD56-negative natural killer cell differentiation in vivo upon KSHV/EBV dual infection

SUMMARY

Herpesvirus infections shape the human natural killer (NK) cell compartment. While Epstein-Barr virus (EBV) expands immature NKG2A⁺ NK cells, human cytomegalovirus (CMV) drives accumulation of adaptive NKG2C⁺ NK cells. Kaposi sarcoma-associated herpesvirus (KSHV) is a close relative of EBV, and both are associated with lymphomas, including primary effusion lymphoma (PEL), which nearly always harbors both viruses. In this study, KSHV dual infection of mice with reconstituted human immune system components leads to the accumulation of CD56⁻CD16⁺CD38⁺CXCR6⁺ NK cells. CD56⁻CD16⁺ NK cells were also more frequently found in KSHV-seropositive Kenyan children. This NK cell subset is poorly cytotoxic against otherwise-NK-cell-susceptible and antibody-opsonized targets. Accordingly, NK cell depletion does not significantly alter KSHV infection in humanized mice. These data suggest that KSHV might escape NK-cell-mediated immune control by driving CD56⁻CD16⁺ NK cell differentiation.

Graphical Abstract

In brief

Caduff et al. utilize a humanized mouse model to characterize NK cell responses upon dual infection with the human tumor viruses KSHV and EBV. They describe an increase of dysfunctional CD56⁻CD16⁺ NK cells with KSHV viremia in vivo and an accumulation of this subset in KSHV-seropositive children.

INTRODUCTION

Two of the seven known human tumor viruses belong to the family of γ-herpesviruses (Bouvard et al., 2009; Parkin, 2006). These are Epstein-Barr virus (EBV) and Kaposi sarcoma-associated herpesvirus (KSHV), associated with around 2% of all malignancies in humans (de Martel et al., 2020). While EBV has causal roles in lymphomas as well as epithelial and rare mesenchymal cell malignancies (Münz, 2019; Shannon-Lowe and Rickinson, 2019), KSHV is associated with B cell lymphoproliferative disorders and the eponymous, endothelial-cell-derived Kaposi sarcoma (KS) (Cesarman et al., 2019; Chang et al., 1994). EBV and KSHV are both thought to drive the malignant transformation of B cells in primary effusion lymphoma (PEL), an aggressive type of large B cell lymphoma with plasmablastic features. PEL cell lines are also the only tumor cells in which KSHV persists in vitro. All cases of PEL are associated with KSHV, and up to 90% also contain EBV (Cesarman, 2014). Accordingly, KSHV persistence is predominantly observed in mice with reconstituted human immune system components (humanized mice) in the presence of EBV, and these dual-infected mice develop PEL-like tumors (McHugh et al., 2017). In good agreement, EBV and KSHV orthologs are also co-transmitted in monkeys (Bruce et al., 2018), KSHV is primarily found in co-existence with EBV in Cameroonian children (Labo et al., 2019), and EBV supports persistence of B cell infection by KSHV in vitro (Faure et al., 2019). Moreover, the high prevalence of EBV in the adult human population (>95%) makes it likely that KSHV co-exists in the presence of EBV in most adult individuals (Dunmire et al., 2018). Thus, any influence of KSHV on human immune compartments should preferably be investigated in the presence of EBV infection.

Despite the strong growth-transforming capacity of EBV and its high prevalence in the human population, very few virus carriers develop associated malignancies (Münz, 2019; Shannon-Lowe and Rickinson, 2019). In most healthy EBV carriers, immune control is thought to prevent the emergence of virus-associated lymphomas, as these can be observed at increased frequencies in patients with immunodeficiencies (Damania and Münz, 2019; Latour and Fischer, 2019; Totonchy and Cesarman, 2016). Natural killer (NK) cells expand during primary EBV infection, control lytic viral replication in humanized mice, and have been identified as crucial components of EBV-specific immune responses by studying primary immunodeficiencies (Azzi et al., 2014; Chijioke et al., 2013; Dunmire et al., 2015; Landtwing et al., 2016; Münz, 2017; Tangye and Latour, 2020; Williams et al., 2005). NK cell activity against target cells is regulated by the balance of germline-encoded activating versus inhibitory receptors, and despite multiple viral evasion strategies, NK cells are particularly potent in restricting herpesvirus infections (Mancini and Vidal, 2020). They contribute to antiviral immunity via cytokine production, such as interferon (IFN)-γ in response to interleukin (IL)-12, and direct killing of infected cells induced by type I IFNs (Vidal et al., 2011). In order to exert these antiviral immune responses, NK cells often require an IL-15-and IL-18-dependent priming step in secondary lymphoid tissues, reminiscent of CD8⁺ T cell priming (Chaix et al., 2008; Lucas et al., 2007). In humans, this activation leads to the stepwise differentiation of CD56^brightCD16⁻ to CD56^dimCD16⁺ NK cells that then successively acquire killer-immunoglobulin-like inhibitory receptors (KIRs) (Björkström et al., 2010a; Ferlazzo et al., 2004). At any differentiation step, NK cells can acquire the lymphocytic senescence marker CD57 and stop expanding (Björkström et al., 2010a). In contrast to KSHV, many NK cell functions and differentiation patterns have been characterized during infections with EBV or the β-herpesvirus cytomegalovirus (CMV). Here, we report that KSHV drives CD56⁻CD16⁺CD38⁺ NK cell differentiation with a partially tissue-resident (CXCR6, CD69) and immunosuppressive phenotype (CD39) in EBV-co-infected humanized mice. CD56⁻CD16⁺ NK cells are also found to be enriched in KSHV-seropositive children in a Kenyan pediatric cohort, supporting our in vivo findings. Although these NK cells retain their cytotoxic machinery, they poorly kill NK cell susceptible targets, even after antibody opsonization.

RESULTS

KSHV infection leads to an accumulation of CD16⁺ NK cells in humanized mice

We investigated the influence of KSHV infection on human NK cells in an in vivo model of persistent KSHV infection developed in our laboratory (McHugh et al., 2017) and based on EBV coinfection of NOD-scid IL2R_γc^−/− (NSG) mice harboring human immune system components. Humanized NSG (huNSG) mice were injected with 10⁶ infectious units (IUs) of KSHV (rKSHV.219) or 10⁵ IUs of EBV (B95-8), both viruses simultaneously or mock-infected with PBS (Figures 1A and S1A). Studies were performed with a lytic replication-deficient EBV strain that lacks BZLF-1 (EBVzko), compared to the wild-type EBV (EBVWT) control, since NK cells mainly control EBV lytically replicating cells (Azzi et al., 2014; Chijioke et al., 2013; Pappworth et al., 2007), and KSHV enhances the lytic activity of EBVWT in huNSG mice (McHugh et al., 2017). Animals were euthanized at 4 weeks post-infection (p.i.) or earlier if they met predetermined humane endpoint criteria, resulting in a significantly reduced survival of dual-infected compared to single- and mock-infected mice (Figure S1B). Spleen and liver sections of infected mice contained CD20⁺ and IRF4⁺/CD20^low cells that were positive for the KSHV latency-associated nuclear antigen (LANA) in areas with a similar pattern of EBV-infected cells on adjacent tissue sections, as determined by the presence of the EBV nuclear antigen (EBNA) 2 and the EBV-encoded small RNAs (EBERs) (Figure 1B). These staining patterns support the previously reported infection of EBV⁺ human B cells with plasma-cell-like phenotype by KSHV in this model (McHugh et al., 2017). The frequency of mice positive for KSHV ORF26 DNA or LANA protein was significantly higher when injected simultaneously with EBV of either strain (Figure 1C; McHugh et al., 2017). Peripheral blood and splenic KSHV DNA loads were similar between EBVWT+KSHV and EBVzko+KSHV mice (Figures 1D and S1C), and EBVzko titers were not affected by KSHV co-infection (Figures 1E and S1D; McHugh et al., 2017). KSHV infection significantly increased the frequency of mice harboring macroscopic tumors when co-infected with EBVWT as compared to EBVWT single-infected animals, but not in mice co-infected with the EBVzko strain (Figure 1F; McHugh et al., 2017). As such, dual infection of mice with KSHV and EBVzko allows the investigation of persistent KSHV infection without the confounding presence of additional KSHV+EBV-driven tumor formation observed during EBVWT co-infection.

Figure 1. KSHV infection leads to the accumulation of CD16⁺ NK cells in huNSG mice.

(A) Experimental outline of EBV and KSHV infection in huNSG mice.

(B) Co-staining of LANA with CD20 and IRF4 on splenic and liver sections with tumor of an EBVWT+KSHV mouse by immunofluorescent (IF) staining. EBER in situ hybridization (ISH) and EBNA2 immunohistochemistry (IHC) staining are shown below. Scale bar: 50 μm.

(C) Frequency of mice injected with KSHV with evidence of viral infection, based on KSHV DNA detection in the blood and spleen and LANA IHC of tissue sections. 15 independent experiments; total number of mice indicated. Fisher’s exact test. (D and E) Mice without quantifiable viral DNA levels are plotted on (D) the x axis or (E) the dashed line indicating the quantification threshold. (D) The area under the curve (AUC) of ORF26 copies/ml at weeks 2, 3, and 4 p.i. was computed to depict the longitudinal peripheral blood KSHV burden. Nine independent experiments with n = 14–22 mice per group. (E) AUC EBV load in international units/ml at weeks 2–4 p.i. depicted for mice of five independent experiments with n = 10–21. (F) Frequency of mice with no tumor, 1 tumor, and ≥2 macroscopic tumors. 15 independent experiments; total number of mice per group indicated. Mann-Whitney U test (MWU) for comparison of tumor score. (G) Total numbers of live, human CD45⁺CD3⁻NKp46⁺ cells/ml peripheral blood before infection (baseline) and at 4 weeks p.i. (sacrifice) are depicted for individual mice of five independent experiments. Baseline and sacrifice NK cell numbers within groups were analyzed by Wilcoxon test.

(H and I) Total numbers of live, human CD45⁺CD3⁻NKp46⁺ cells per (H) spleen or (I) liver. Median, IQR. Composite data from (H) nine independent experiments with n = 20–30 mice per group and (I) two independent experiments with n = 8–12 mice per group.

(J) Frequency of CD16⁺ within human peripheral blood CD45⁺CD3⁻NKp46⁺ cells over time in mock (n = 5), EBVzko (n = 5), and EBVzko+KSHV (n = 6) mice. Statistical analysis was performed with mixed-effects model (REML) followed by Fisher’s LSD test. False discovery rate (FDR)-adjusted p (q) values, mean ± SEM.

(K and L) Frequency of human CD45⁺Lin⁻ (CD3, CD19, CD14, CD4) NKp46⁺ (K) splenocytes or (L) liver cells positive for CD16 of mice from four and two independent experiments, respectively, relative to the mean of the respective EBVzko group.

Mean ± SEM, MWU. *p < 0.05, **p < 0.01, ***p < 0.001. Unless otherwise stated, statistical analysis was performed using Kruskal-Wallis (KW) test followed by Dunn’s multiple comparison test with q values indicated. See also Figure S1.

The majority of human NK cells in huNSG mice express the natural cytotoxicity receptor NKp46 (Chijioke et al., 2013; Strowig et al., 2010). Upon infection, peripheral blood human NKp46⁺CD3⁻ NK cell numbers increased, but splenic or hepatic NK cell counts 4 weeks p.i. were similar between EBV single- and EBV+KSHV co-infected animals (Figures 1G–1I). However, we documented an increase in the frequency of CD16⁺ NK cells in the blood of dual-infected mice over time (Figure 1J), and a significantly higher proportion of this subset was also found in spleen and liver of dual-infected mice compared to EBVzko single-infected animals (Figures 1K and 1L). Thus, our data suggest that infection with KSHV leads to an accumulation of the more differentiated CD16⁺ NK cell subset.

CD56⁻CD16⁺ NK cells are enriched in the NK cell compartment of KSHV-infected humanized mice as well as Kenyan KSHV-seropositive children

In humans, the expression of CD56 and CD16 distinguishes two major subsets: the mainly cytokine-producing CD56^brightCD16⁻NK cells and the more cytotoxic CD56^dimCD16⁺ subset that constitutes the majority of NK cells in human peripheral blood. In naive huNSG mice, the majority of NK cells are CD56⁺CD16⁻, the subset that is considered less mature than its CD16⁺ counterpart (Strowig et al., 2010). Interestingly, the significant increase in CD16⁺ NK cell frequency in EBVz-ko+KSHV co-infected mice was due to an increase in the frequency of a non-classical CD56⁻CD16⁺ NK cell population at the expense of CD56⁺CD16⁻ cells (Figure 2A). We observed a gradual increase in the frequency of peripheral blood CD56⁻CD16⁺ NK cells over time in EBVzko+KSHV co-infected, but not EBVzko single- or mock-infected, animals (Figure 2B). This difference in the NK cell subset distribution was also observed in the spleen and liver of EBVzko+KSHV co-infected, compared to EBVzko single-infected, animals (Figures 2C and 2D). Within dual-infected animals, KSHV titers positively correlated with the relative abundance of splenic CD56⁻CD16⁺ NK cells, and a negative correlation was observed between splenic KSHV loads and the frequency of CD56⁺CD16⁻ NK cells (Figure 2E). EBV viral loads did not positively correlate with CD56⁻CD16⁺ NK cell frequency (Figure S1E). Concurrently, KSHV co-infected mice had significantly higher serum levels of IL-15, IL-18, and IL-27 (Figure 2F), but not IL-12p70, IL-2, FLT3L, and type I IFNs (Figures S1F and S1G).

Figure 2. CD56⁻CD16⁺ NK cells are enriched in the NK cell compartment of KSHV-infected huNSG mice, as well as in Kenyan KSHV-sero-positive children.

(A) Representative flow cytometry dot plots of human CD45⁺NKp46⁺Lin⁻ splenocytes of EBVzko and EBVzko+KSHV mice.

(B) Frequency of peripheral blood CD56-and CD16-stained NK cell subsets over time in mock (n = 5), EBVzko (n = 5), and EBVzko+KSHV (n = 6) mice. REML followed by Fisher’s LSD test, q values, mean ± SEM. (C and D) Frequency of CD56-and CD16-stained NK cell subsets among splenic (C) or hepatic (D) CD45⁺CD3⁻NKp46⁺ cells of mice from four and two independent experiments, respectively, relative to the mean of the respective EBVzko group. Mean ± SEM, MWU.

(E) Correlations between splenic KSHV DNA load and the frequency of CD56⁺CD16⁻ (red) and CD56⁻CD16⁺ (yellow) NK cells are depicted for EBVzko+KSHV co-infected mice from four independent experiments. Solid lines represent trend lines obtained by linear regression, and shaded areas indicate 95% confidence interval (CI) of each trend line. Mice with values below 0.1 ORF26 copies per 10⁶ cells were excluded.

(F) Cytokine concentration was measured in the serum of EBVzko (n = 15) and EBVzko+KSHV (n = 16) mice at 4 weeks p.i. The median (IQR) is shown with minimum and maximum ranges as whiskers. Composite data are from four independent experiments, MWU. Dashed lines indicate lower level of quantification (LLOQ). (G) Frequency of CD56⁻CD16⁺ NK cells in KSHV-seropositive and KSHV-negative (KSHVab + and −) pediatric healthy donors (HDs). Mean ± SD. MWU.

(H) Correlation between CD56⁻CD16⁺ NK cell frequencies and serum IL-18 in pediatric HDs and eBL patients.

*p < 0.05, **p < 0.01, ***p < 0.001. r_S, Spearman’s correlation. See also Figure S1.

In healthy adults, CD56⁻CD16⁺ NK cells are present in very limited numbers (Björkström et al., 2010b). Among healthy Kenyan children, CD56⁻CD16⁺ NK cells were significantly more frequent in KSHV-seropositive compared to seronegative donors (Figure 2G). However, in Kenyan pediatric patients with endemic Burkitt lymphoma (eBL), CD56⁻CD16⁺ NK cell frequency was elevated and not further increased with seropositivity to KSHV (Figure S1H). Nevertheless, CD56⁻CD16⁺ NK cell frequencies correlated with IL-18 serum levels across all tested groups (Figures 2H and S1I). These data suggest that KSHV infection drives CD56⁻CD16⁺ NK cell accumulation in both huNSG mice and the investigated pediatric cohorts, potentially via IL-15, IL-18 and IL-27.

KSHV infection drives CD56⁻CD16⁺ NK cell differentiation with expression of CD38, CD39, CD69, and CXCR6 in humanized mice

In order to characterize the accumulating CD56⁻CD16⁺ NK cells in more detail, we assessed their maturation and activation status in spleen and liver of EBVzko+KSHV-infected huNSG mice by flow cytometry. Nearly all NKp46⁺Lir⁻ cells co-expressed the pan-NK cell marker CD7, which is not found on mature myeloid cells (Milush et al., 2009), and only a minority of Lir⁻CD7⁺ cells did not express NKp46 (Figures S2A and S2B). After manual gating on human CD45⁺CD7⁺Lir⁻ lymphocytes, uniform manifold approximation and projection (UMAP) was used for dimensionality reduction and visualization of cells (Figures 3A and 3B). CD16⁺ NK cell subsets expressed less NKp46 as well as NKG2A and CD94 (Figure 3C), the two components of an inhibitory C-type lectin receptor, described to be highly expressed by CD56^beight NK cells and to decrease during differentiation (Björkström et al., 2010a; Di Vito et al., 2019; Yu et al., 2010) (Figures 3C, S2C, and S3A). Furthermore, CD16⁺ NK cells were largely negative for early NK differentiation and innate lymphoid cell markers CD117 and CD127 (Figures 3D, S2D, and S3B). Inversely, the expression of KIRs and CD57, which are progressively acquired with differentiation, was confined to the CD16⁺ subsets and was similar between CD56⁺CD16⁺ and CD56⁻CD16⁺ cells.

Figure 3. KSHV infection drives CD56⁻CD16⁺ NK cell differentiation with expression of CD38, CD39, CD69, and CXCR6 in huNSG mice.

(A) UMAP visualization of single, live, human CD45⁺Lin⁻ (CD3, CD19, CD14, CD4) CD7⁺ cells isolated at 4 weeks p.i. from spleen and liver of mock-, EBVzko-, and EBVzko+KSHV-infected mice depicted as contour plot.

(B) UMAP plot depicting relative marker expression from samples in (A). (C) Heatmap depicting median fluorescent intensity (MFI) of surface markers on CD56-and CD16-stained NK cell subsets (pre-gated on human CD45⁺NKp46⁺Lin⁻) and on human CD45⁻ cells of EBVzko+KSHV co-infected animals.

(D) Heatmap depicting the median frequency of cells positive for indicated surface markers among different CD56-and CD16-stained NK cell subsets. In (C) and here, composite data are from 1–4 independent experiments with n = 6–20 animals. Data are normalized per column. Statistical analysis was performed by multiple comparison tests following paired one-way ANOVA or Friedmann’s test (q values). See also Figures S3A and S3B.

(E and F) Expression of (E) CD38 and CD2 or (F) CXCR6 and CD62L among splenic and hepatic NK cells. Colors in the pie charts indicate CD56/CD16 subsets. Pie arc colors depict the expression of (E) CD38 and CD2 or (F) CXCR6 and CD62L. In (E), scatterplots depict subset frequencies within NKp46⁺ cells of individual mice (symbol color: black, spleen; clear, liver). Median.

(G) Frequencies of CXCR6⁺ splenic or hepatic NK cells.

(H and I) Frequencies of cells expressing (H) CD69 or (I) CD39 among CXCR6⁺ and CXCR6⁻ NK cells from EBVzko+KSHV mice. Wilcoxon test.

In (G)–(I), three independent experiments. Median (IQR). In (C)–(I), NK cells were pre-gated on huCD45⁺, live, Lin⁻ (CD3, CD19, CD14, CD4) NKp46⁺ cells. *p < 0.05, **p < 0.01, ***p < 0.001. Unless otherwise stated, statistical analysis was performed by Dunn’s test following KW test (q values).

See also Figures S2 and S3.

CD56⁻CD16⁺ NK cells less frequently expressed CD2, a costimulatory receptor important for the signaling of NKp46 and CD16 (Bryceson et al., 2006; Liu et al., 2016), compared to CD56⁺CD16⁺ NK cells, whereas nearly all CD56⁻CD16⁺ NK cells expressed the activation marker CD38 (Figure 3D). When comparing NK cells from different experimental groups, we observed a significant increase in CD56⁻CD16⁺CD38⁺ cells in EBVzko+KSHV co-infected compared to control mice and a concomitant decrease of CD56⁺CD16⁻CD2⁺CD38⁻ cells (Figures 3E and S3C). Furthermore, dual infection with KSHV led to the accumulation of splenic and hepatic NK cells expressing the tissue residency marker CXCR6 (Dogra et al., 2020; Harmon et al., 2016), frequently co-expressing CD69 (Figures 3F–3H, S3D, and S3E). CXCR6 expression further correlated with KSHV titers (Figure S3F). Lastly, CXCR6⁺ NK cells preferentially co-expressed CD39 (Figure 3I), a cell-surface ATP ectonucleotidase associated with immunosuppression and cell exhaustion (Canale et al., 2018; Tøndell et al., 2020; Wu et al., 2020; Zheng et al., 2020). Thus, KSHV infection seems to promote the expansion of activated, differentiated, and potentially immunosuppressive NK cells with hallmarks of tissue residency in vivo.

KSHV infection increases NK cell cytotoxicity, but CD56⁻CD16⁺ NK cells kill poorly even after target cell opsonization

Next, we addressed NK cell functions in KSHV-infected huNSG mice, as hyporesponsiveness to target cells and cytokine stimulation has been described for CD56⁻CD16⁺ NK cells (Björkström et al., 2010b). Proliferating (Ki-67⁺) splenic NK cells at 4 weeks p.i. were significantly more frequent in EBVzko+KSHV co-infected, compared to EBVzko and mock, animals and were preferentially CD56⁺CD16⁻CD57⁻ (Figures 4A and S4A). In line with a more mature phenotype, CD16⁺ NK cell subsets more frequently contained effector molecules, granzyme B, and/or perforin, compared to the CD16⁻ counterparts, with the highest perforin levels found in CD56⁻CD16⁺ cells (Figures 4B and S4B). Overall, the frequency of perforin-/granzyme-B-expressing NK cells was highest in EBVzko+KSHV mice (Figure 4C), indicating increased activation and “priming” of NK cells upon dual infection. Accordingly, NK cells from EBVzko+KSHV co-infected mice degranulated significantly more frequently against K562 cells compared to NK cells from control mice (Figures 4D, 4E, and S4C). The majority of degranulation occurred, however, among CD56⁺CD16⁻ and not in the accumulating CD56⁻CD16⁺ NK cell subset (Figure 4F). Although shedding of CD16 upon stimulation, as previously described (Goodier et al., 2016; Romee et al., 2013), may have caused underestimation of CD56⁻CD16⁺ NK cell degranulation, the overall poor degranulation of CD56⁻ subsets against K562 was insufficient to account for the superior performance of NK cells from EBVzko+KSHV mice (Figure S4D). Besides pronounced shedding of CD16 upon contact with opsonized cells (Figures 4G and 4H), degranulation mediated by antibody-dependent cellular cytotoxicity (ADCC) was only slightly increased in EBVzko+KSHV-infected mice (Figures 4D and S4C). Similar to the superior degranulation of CD56⁺ subsets against K562 targets, CD56⁺ cells more readily produced IFN-γ upon cytokine stimulation (Figure S4E). Furthermore, serum levels of the potentially NK-cell-derived cytokines IFN-γ and tumor necrosis factor alpha (TNF-α) were increased in both EBVzko- and EBVzko+KSHV-infected, compared to mock-infected, mice (Figures 4I and S4F). Thus, enhanced NK cell activation during KSHV co-infection leads to the accumulation of hyporesponsive CD56⁻CD16⁺ NK cells and the increased proliferation, cytokine production, and cytotoxicity of a diminished number of CD56⁺CD16⁻ NK cells.

Figure 4. KSHV infection increases NK cell cytotoxicity, but CD56⁻CD16⁺ NK cells kill poorly even after target cell opsonization.

(A) Frequency of splenic Ki-67⁺ NK cells of individual mice from two independent experiments. Median. Heatmap depicts median frequencies upon sub-gating on CD56- and CD16-stained NK cell subsets.

(B) Representative MFI histograms of granzyme B and perforin expression within huNSG splenic NK cell subsets isolated at 4 weeks p.i. compared to NK cells from human peripheral blood mononuclear cells (PBMCs) and fluorescence minus one (FMO) controls. Scatterplots depict frequencies of granzyme B⁺ and perforin⁺ splenocytes among subsets from EBVzko+KSHV mice of two independent experiments.

(C) Frequency of NK cells dual positive for granzyme B and perforin. Median (IQR). In (D)–(H), ex vivo degranulation assays at 4 weeks p.i. are shown. (D) Representative contour plots of the CD107a staining upon co-culture of huNSG splenocytes (top) or human PBMCs (bottom) with K562 cells, Raji cells, Raji cells + rituximab (RTX), or medium (unstimulated control). HuNSG NK cells were pre-gated on live huCD45⁺Lin⁻ NKp46⁺ cells; human PBMCs were gated on live huCD45⁺Lin⁻CD7⁺ cells.

(E and F) Specific degranulation of NK cells from (E) different experimental groups and (F) upon sub-gating on NK populations, as measured by the frequency of CD107a⁺ cells upon contact with K562 minus background (unstimulated control). Two independent experiments. (E) Median. (F) Mean +SD.

(G) Representative flow cytometry contour plots of CD56 and CD16 expression upon co-culture as in (D). CD107a⁺ cells are overlaid as red dots.

(H) Frequency of CD16⁺ NK cells (Lin⁻NKp46⁺) after co-culture as in (D). Connected dots represent individual mice. Mean is shown in red. Friedman’s test (q values).

(I) Serum cytokine concentration at 4 weeks p.i. of mock (n = 9), EBVzko (n = 14), and EBVzko+KSHV (n = 11) mice from five independent experiments (heatmap values scaled per row). (J) Experimental outline of NK cell depletion experiments (n = 2) and frequency of peripheral blood huCD45⁺CD3⁻NKp46⁺ cells over time. Mean ± SEM.

(K and L) AUC of (K) peripheral blood KSHV ORF26 DNA levels (weeks 2–4 p.i.) and (L) at 4 weeks p.i. in the spleen. Median (IQR), MWU. (M) AUC (weeks 2–4 p.i.) of peripheral blood EBV international units/ml. Median (IQR), MWU. *p < 0.05, **p < 0.01, ***p < 0.001. Unless otherwise stated, statistical analysis was performed by Dunn’s test following KW test (q values). See also Figure S4.

Accordingly, NK cell depletion with an NKp46-specific antibody (Chijioke et al., 2013; Landtwing et al., 2016) did not significantly affect KSHV viral burden, KSHV persistence rate, or tumor formation in huNSG mice (Figures 4J–4M, and S4G–S4I). Altogether, KSHV infection differentiates NK cells toward the CD56⁻CD16⁺ subset that, despite its high cytotoxic molecule content, seems to be maintained in a reduced functional state, and the NK cell compartment as a whole is unable to meaningfully control KSHV.

DISCUSSION

In this study, we have investigated the effect of persistent KSHV infection on the NK cell compartment in humanized mice. We observed an accumulation of CD56⁻CD16⁺CD38⁺CXCR6⁺CD69⁺CD39⁺ NK cells, which correlated with KSHV viral burden. Likewise, Kenyan KSHV-seropositive children were found to carry higher frequencies of CD56⁻CD16⁺ NK cells. This is a quite unique observation among herpesviruses, since infection with the α-herpesvirus herpes simplex virus (HSV) does not seem to change the composition of the NK cell repertoire (Björkström et al., 2011), the β-herpesvirus CMV drives the accumulation of CD56⁺CD16⁺NKG2C⁺KIR⁺ adaptive NK cells (Gumá et al., 2004; Smith et al., 2020), and the γ-herpesvirus EBV expands CD56⁺CD16^+/−NKG2A⁺KIR⁻ NK cells (Azzi et al., 2014). The functional relevance of these compositional changes in the NK cell repertoire for the immune control of viruses, however, often remains unclear.

Along these lines, it had originally been suggested that NK cells can kill KSHV-infected cells (Lebbé et al., 1997; Matthews et al., 2011; Sirianni et al., 2002). However, NK cell cytotoxicity is reduced in patients with KSHV-associated malignancies, in part via downregulation of activating NK cell receptors and increased expression of inhibitory receptors (Beldi-Ferchiou et al., 2016; Dupuy et al., 2012; Sirianni et al., 2005). While many immune escape mechanisms have been described for CMV (Mancini and Vidal, 2020), NK cell responses are protective against infection with this β-herpesvirus in preclinical in vivo models. In contrast, we report that the accumulation of poorly cytotoxic CD56⁻CD16⁺ NK cells during KSHV infection is associated with a lack of NK-cell-mediated immune control of this γ-herpesvirus infection in vivo. We show that antibody-mediated NK cell depletion, which compromises immune control of EBVWT infection in the same humanized mouse model (Chijioke et al., 2013; Landtwing et al., 2016), does not lead to increased KSHV titers or more frequent persistent KSHV infection. Thus, the decreased in vitro cytotoxicity of CD56⁻CD16⁺CD38⁺CXCR6⁺ NK cells that accumulate during KSHV infection in humanized mice is in line with their inability to control this γ-herpesvirus in vivo.

In good agreement, CD56⁻CD16⁺ NK cells display compromised functions during other chronic viral infections despite phenotypically resembling CD56⁺CD16⁺ NK cells in many aspects (Björkström et al., 2010b; Forconi et al., 2018,2020; Milush et al., 2013; Müller-Durovic et al., 2019; Voigt et al., 2018). CD56⁻CD16⁺ NK cell numbers increase with HIV-viremia and decrease upon successful antiretroviral therapy (Alter et al., 2005; Mavilio et al., 2005; Milush et al., 2013). CD56⁻CD16⁺ NK cells also accumulate in elderly people co-infected with EBV and CMV (Müller-Durovic et al., 2019) and in pediatric patients with EBV⁺ post-transplant lymphoproliferative disease (Wiesmayr et al., 2012) or eBL (Forconi et al., 2018). The latter association might, however, also result from holoendemic malaria exposure (Forconi et al., 2018), which is itself associated with KSHV seropositivity (Nalwoga et al., 2015; Oluoch et al., 2020). While it is difficult to control and account for multiple pathogens or underlying health issues during natural infection in humans where exposure cannot be assigned, studies in humanized mice allow the investigation of single parameters or a combination thereof, facilitating the interpretation of likely causes and consequences of CD56⁻CD16⁺ NK cell accumulation. At the same time, de novo infections in huNSG mice may not allow the study of long-term effects of chronic infection or tumorigenesis.

A pro-inflammatory environment common to all these conditions, either by one virus or by a combination of pathogens, might drive CD56⁻CD16⁺ NK cell differentiation. The cytokines IL-15, IL-18, and IL-27, identified to be elevated upon KSHV and EBV co-infection in this study, are good candidates to drive this differentiation, and their role should be investigated in future studies. Moreover, NK cell differentiation by KSHV is associated with upregulation of molecules that facilitate tissue residency, like CXCR6 and CD69 (Freud et al., 2017), accompanied by immunosuppressive CD39 expression. Thus, KSHV might escape NK-cell-mediated immune control by reducing cytotoxicity and patrolling patterns of these innate lymphocytes, with potential relevance in the development of KS and PEL, which predominantly occur in the context of HIV co-infection.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by Prof. Christian Münz, PhD (Christian.muenz@uzh.ch).

Materials availability

This study did not generate new unique reagents.

Data and code availability

This study did not generate datasets or codes.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human tissue and blood samples

The ethical approval for the use of human fetal liver tissue as well as peripheral blood samples of healthy adult volunteers were obtained from the cantonal ethical committee of Zurich, Switzerland (protocol no. KEK-StV-Nr.19/08). Studies involving peripheral blood of healthy children and pediatric endemic Burkitt lymphoma (eBL) patients in Kenya, described previously in Forconi et al. (2018) and Oluoch et al. (2020), were reviewed and approved by the Scientific and Ethics Review Unit at the Kenya Medical Research Institute, Kenya, and the Institutional Review Board at the University of Massachusetts Medical School, USA. Written informed consent was obtained from parents of pediatric study participants prior to study inclusion. Samples from 14 children with eBL (median age 5.8 years; 64% male) and 11 children without eBL (median age 6 years; 72% male) were used for this study. All subjects were HIV-negative and born to HIV-negative mothers.

Humanized mouse generation and infection

NOD-scid γc^−/− mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ or NSG) and HLA-A2 transgenic NSG (NOD.Cg-Mcph1^{Tg(HLA-A2.1)1Enge} Prkdc^scid Il2rg^tm1Wjl/SzJ or NSG-A2) mice were purchased from The Jackson Laboratory and bred and maintained at the Institute of Experimental Immunology, University of Zurich. At 1-5 days of age and 5-7 hours after sub-lethal irradiation (1Gy), mice were injected intrahepatically with 1-3 × 10⁵ human fetal liver (HFL)-derived hematopoietic progenitor cells (HPCs). HFL tissue was obtained from Advanced Bioscience Resources, and HPCs were isolated by magnetic cell sorting for CD34 positive cells according to the manufacturer’s instructions (Miltenyi Biotec) and as previously described (Strowig et al., 2009; White et al., 2012). Three months after HPC injection and again one week before infection, the reconstitution of human cells was assessed in the peripheral blood of mice by flow cytometry for human CD45 (huCD45), huCD3, huCD19, huCD4, huCD8 and huNKp46 as previously described (McHugh et al., 2020). Three to six months after engraftment (average: 4.3 months), mice were injected intraperitoneally (i.p.) with 10⁶ HEK293T-lnfectious Units (IU) of KSHV (rKSHV.219), 10⁵ Raji IU of EBV (B95-8), simultaneously injected with both viruses or mock-infected with PBS. Apart from tail blood vessel puncture, mice were not involved in any procedures prior to the infection with EBV and KSHV, with the exception of animals that were part of NK cell depletion experiments. These mice received i.p. injections of 100μg purified anti-NKp46 antibody (clone BAB281) in PBS, or mock-injections with PBS, on three consecutive days (total of 30μg per mouse) as previously described (Chijioke et al., 2013). The following day, after assessing the depletion efficiency in peripheral blood by flow cytometry, mice were infected with EBV and KSHV as described. Mice were euthanized at four weeks p.i. or earlier if the predetermined humane endpoint criteria of the laboratory’s animal welfare protocol were met that included weight loss (loss of 20% of body weight from maximum weight or 15% persisting for two days) and a scoring for signs of general health and distress (e.g., hunched posture, piloerection, inactivity, grimace). The frequency of macroscopic tumors was assessed by a score: no tumors observed = 0; one tumor observed = 1; multiple tumors observed = 2. EBV-inoculated mice without detectable EBV (BamHI W fragment) DNA in the blood or spleen, and no positive signal for EBNA2 staining or EBER ISH in FFPE tissue sections were considered non-infected and excluded from this study. Similarly, EBV+KSHV-inoculated mice without detectable KSHV (ORF26) DNA in the blood or spleen, and no positive signal for LANA in FFPE tissue sections were considered KSHV non-infected and excluded.

This study includes data from mice reconstituted with CD34⁺ HPCs isolated from a total of 15 different HFL donors; 13 reconstituted NSG cohorts with 190 animals and 2 reconstituted NSG-A2 cohorts with 36 animals. Individual experiments were performed with a single cohort consisting of mice reconstituted with cells from the same HFL. Mice were distributed to experimental groups under consideration of the sex and frequencies of human cells in the peripheral blood, ensuring a similar sex-ratio and overall similar human immune cell reconstitution between the groups. Investigators were not blinded regarding the grouping of mice. Experiments comprise a total of 137 female mice (Mean weight ± SD: 20.3g ± 1.7g) and 89 male mice (26.3 ± 3.2g) of 20.2 ± 4.7 weeks of age (Mean ± SD; range: 13-29 weeks). The median frequency (and IQR: 25^th, 75^th percentile) of huCD45⁺ cells of peripheral blood lymphocytes before infection was 82% (70.5%, 87.1%), huCD3⁺ T cells of huCD45⁺ lymphocytes: 33.6% (24.3%, 49.2%), huCD19⁺ B cells of huCD45⁺ lymphocytes: 58.1% (42.8%, 67.6%), huCD4⁺ cells of huCD3⁺ huCD45⁺ lymphocytes: 72% (65.4%, 79%), huCD8⁺ cells of huCD3⁺ huCD45⁺ lymphocytes: 24.1% (19%, 30.9%), CD3⁻ NKp46⁺ NK cells of human huCD45⁺ lymphocytes: 1.9% (1.4%, 2.7%) (Figure S1A). Animals were housed in groups and kept singly only in exceptional circumstances (e.g., due to aggressive behavior or absence of male littermates). Animal protocols were reviewed and approved by the veterinary office of the canton of Zurich, Switzerland (116/2008, 148/2011, 209/2014, 159/2017).

METHOD DETAILS

Recombinant EBV and KSHV

EBV B95-8-GFP (EBVwt, p2089) and EBV B95-8 ΔBZLF1 -GFP (EBVzko) were produced in HEK293 cells as described previously (Delecluse et al., 1998; Feederle et al., 2000). Recombinant KSHV (rKSHV.219) was produced in Vero cells as previously described (Vieira and O’Hearn, 2004), or in latently-infected BJAB cells (BrK.219) (Kati et al., 2013) by inducing the lytic cycle using a combination of anti-IgM antibody (0.625 μg/ml, Southern Biotech) and TPA (12-O-Tetradecanoylphorbol 13-acetate, 0.05 μg/ml, Sigma-Aldrich) adapted from the protocol of Kati et al. (2015). KSHV produced in Vero cells was kindly provided by Prof D. J. Blackbourn (Guildford, UK) and BrK.219 cells were kindly gifted by Prof T. F. Schulz (Hannover, Germany). Titers of virus concentrates were determined by flow cytometric analysis of GFP-positive cells 48 hours after in vitro infection of Raji cells by EBV or HEK293T cells by KSHV on a FACSCanto II (BD Biosciences), as previously described (Antsiferova et al., 2014).

Quantification of EBV and KSHV DNA

For humanized mouse samples, total DNA from whole blood or serum was extracted using the NucliSENS EasyMag System (bio-Mérieux) according to the manufacturer’s protocol. DNA from spleen tissue was isolated using the DNeasy Blood & Tissue Kit (QIAGEN). Quantification of EBV BamHI W DNA was performed by TaqMan (Applied Biosystems) real-time PCR on an ABI Prism 7700 Sequence detector (Applied Biosystems) with the following primer and FAM/TAMRA-labeled probe sequence (modified from Berger et al., 2001): 5′-CTTCTCAGTCCAGCGCGTTT-3′, 5′-CAGTGGTCCCCCTCCCTAGA-3′, 5′-(FAM)-CGTAAGCCAGACAGCAGCCAATTGTCAG-(TAMRA)-3′. KSHV ORF26 DNA was measured by TaqMan (Applied Biosystems) quantitative PCR on a C1000 Touch CFX384 Real-Time platform (Bio-Rad, Hercules, CA, USA) as previously described (McHugh et al., 2017) with the following primer and probe sequence (modified from Tedeschi et al., 2001): 5′-GCTCGAATCCAACGGATTTG-3′, 5′-AATAGCGTGCCCCAGTTGC-3, 5′-(FAM)-TTCCCCATGGTCGTGCCTC-(BHQ-1)-3′. Samples were analyzed in duplicates or triplicates.

For human samples, peripheral blood DNA was extracted using the DNeasy Blood & Tissue Kit (QIAGEN) according to the manufacturer’s protocol and frozen at −20°C until use. EBV DNA was quantified by TaqMan (Applied Biosystems) real-time PCR targeting the BALF5 gene and normalized to β-actin, as described by Forconi et al. (2018) and Moormann et al. (2005). The previously determined threshold of 3 log EBV copies / μg DNA was used to group samples into the 2 groups: no / low EBV and high EBV.

KSHV serology by multiplex beads assay

We used 7 KSHV recombinant open reading frames (ORFs): ORF59, ORF73/ latency-associated nuclear antigen (LANA), ORF65, ORF61, ORF38, K8.1, and K5. Bio-plex COOH carboxylated nonmagnetic beads (1.25 × 10⁷ microspheres/mL) were coupled to corresponding antigens at 100 μg/500 μL based on manufacturer’s protocol, and antibody levels for KSHV, EBV, and malaria measured by BioPlex 200 Multianalyte Analyzer (Bio-Rad Laboratories) as previously described in Oluoch et al. (2020). In order to determine inter/intraassay variability, negative and positive control samples were included on each plate in duplicates. A comparison of the mean and standard deviation of duplicates within and between/across plates was used to determine the percent relative standard deviation (%RSD), which showed a range of 2%–15% for KSHV antigens. To determine individual seropositivity cutoffs, we used the receiver operating characteristic (ROC) curve analysis (Greiner et al., 2000) of the 7 KSHV antigens based on median fluorescence intensities (MFI) of 11 Africans and 15 North Americans adults with high and low risks of KSHV infection, respectively.

Cell isolation

Spleen tissue was manually dissociated, filtered through 70-μm cell strainer, and subjected to Ficoll gradient centrifugation (Ficoll Paque Premiem, GE Healthcare). Liver tissue was digested for 30min at 37°C in buffer containing DNase (0.4mg/ml) and Collagenase D (20μg/ml) (Roche), followed by addition of 0.5M EDTA, filtering and separation of mononuclear cells on Percoll gradients. Whole blood was collected by tail-vein and terminal cardiac puncture in tubes containing heparin or EDTA. Prior to erythrocyte lysis by ACK lysis buffer, the white blood cell count in whole blood was determined with a hematology analyzer (Beckman Coulter AcT Diff Analyzer). Human PBMCs were isolated from Lithium-Heparin vacutainers by Ficoll gradient centrifugation within 2 hours of the blood draw and cryopreserved until use, as described by Forconi et al. (2018).

Flow cytometric analysis

All antibodies are listed in the Key resource table. Dead cells were excluded using Fixable Viability dyes (Near-IR and Aqua, Biolegend). For surface stainings, single cell suspensions were incubated with fluorophore-conjugated antibodies for 30min at 4°C, washed and fixed with 1% PFA. For intracellular stainings, cells were fixed and permeabilized with the Cytofix/Cytoperm Kit (BD Bioscience) after surface antigen staining had been performed. Intranuclear stainings were performed after fixation and permeabilization with the Foxp3 staining kit (eBioscience) according to the manufacturer’s instructions. Acquisition and compensation was performed on BD FACS Canto II, BD LSR II Fortessa and BD FACSymphony A5 flow cytometers. Data were exported and analyzed using FlowJo software (version 10, TreeStar Inc). For UMAP visualizations, FCS files of compensated, quality-controlled, single, live, human CD45⁺CD7⁺Lin⁻ (CD3, CD4, CD19, CD14) cells were exported from FlowJo into the R environment (R Core Team, 2020) where data were arcsine-transformed using transformation cofactors determined in Cytobank (https://www.cytobank.org), followed by percentile normalization based on the workflow from Hartmann et al. (2016) and UMAP dimensionality reduction (Mclnnes et al., 2018; https://github.com/lmcinnes/umap) with the following parameters: n_neighbors = 20, n_components = 2, metric = ”euclidean,”, n_epochs = 250 and min_dist = 0.3. Heatmaps, graphs and pie charts were generated in R, using Prism (GraphPad Software) or SPICE software (Roederer et al., 2011).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BUV395 mouse anti-human CD45 (Clone: HI30)	BD Bioscience	Cat#563792; RRID:AB_2744400
BUV563 mouse anti-human CD56 (Clone: NCAM16.2)	BD Bioscience	Cat#565705; RRID:AB_2744431
BUV737 mouse anti-human CD16 (Clone: 3G8)	BD Bioscience	Cat#612786; RRID:AB_2833077
BUV805 mouse anti-human CD69 (Clone: FN50)	BD Bioscience	Cat#748763; RRID:AB_2857327
PB mouse anti-human CD45 (Clone: HI30)	BioLegend	Cat#304029; RRID:AB_2174123
BV421 mouse-anti human CD69 (Clone: FN50)	Biolegend	Cat# 310929 RRID:AB_10933255
BV421 mouse anti-human Perforin (Clone: dG9)	Biolegend	Cat#308122; RRID:AB_2566204
BV421 mouse anti-human CD94 (Clone: HP-3D9)	BD Bioscience	Cat#743948; RRID:AB_2741870
BV421 mouse anti-human CD3 (Clone: UCHT1)	Biolegend	Cat#300434 RRID:AB_10962690
BV421 mouse anti-human CD14 (Clone: HCD14)	Biolegend	Cat#325628 RRID:AB_2563296
BV510 mouse anti-human CD16 (Clone 3G8)	Biolegend	Cat# 302048; RRID: AB_2562085
BV510 mouse anti-human CD3 (Clone: UCHT1)	Biolegend	Cat#300448; RRID:AB_2563468
BV510 mouse anti-human CD4 (Clone: OKT4)	Biolegend	Cat#317444; RRID:AB_2561866
BV510 mouse anti-human CD19 (Clone: HIB19)	Biolegend	Cat#302242; RRID:AB_2561668
BV510 mouse anti-human CD14 (Clone: M5E2)	Biolegend	Cat#301842; RRID:AB_2561946
BV605 mouse anti-human Ki-67 (Clone: Ki-67)	Biolegend	Cat#350522; RRID:AB_2563863
BV605 mouse anti-human CD117 (Clone: 104D2)	Biolegend	Cat#313218; RRID:AB_2562025
BV605 mouse anti-human CD39 (Clone A1)	Biolegend	Cat#328235; RRID: AB_2750429
BV605 mouse anti-human CD4 (Clone: OKT4)	Biolegend	Cat#317438; RRID:AB_11218995
BV711 mouse anti-human CD7 (Clone: M-T701)	BD Bioscience	Cat#564018; RRID:AB_2738544
BV711 mouse anti-human Ki-67 (Clone: Ki-67)	Biolegend	Cat#350516; RRID:AB_2563861
BV786 mouse anti-human CD161 (Clone: DX12)	BD Bioscience	Cat#744096; RRID:AB_2741990
BV786 mouse anti-human IFN-γ (Clone: 4S.B3)	BD Bioscience	Cat#563731; RRID:AB_2738391
BV785 mouse anti-human CD3 (Clone: OKT3)	Biolegend	Cat#317330; RRID:AB_2563507
FITC mouse anti-human CD94 (Clone: DX22)	Biolegend	Cat#305504; RRID:AB_314534
FITC mouse anti-human CD2 (Clone: RPA-2.10)	Biolegend	Cat#300206; RRID:AB_314030
FITC mouse anti-human CD107a (Clone: H4A3)	BD Bioscience	Cat#555800; RRID:AB_396134
FITC mouse anti-human CD158a,h,g (Clone: HP-MA4)	Biolegend	Cat#339504; RRID:AB_2130378
FITC mouse anti-human CD158b (Clone: DX27)	Biolegend	Cat#312604; RRID:AB_2296486
FITC mouse anti-human CD3 (Clone: UCHT1)	Biolegend	Cat#300406; RRID: AB_314060
PerCp mouse anti-human CD8 (Clone: SK1)	Biolegend	Cat#344708; RRID:AB_1967149
PE mouse anti-human CD159a/NKG2A (Clone: Z199)	Beckman Coulter	Cat#IM3291U; RRID:AB_10643228
PE mouse anti-human CD328/Siglec-7 (Clone: 6-434)	Biolegend	Cat#339204; RRID:AB_1501160
PE mouse anti-human CD19 (Clone: HIB19)	Biolegend	Cat#302208; RRID:AB_314238
PE/Dazzle 594 mouse anti-human CD186/CXCR6 (Clone: K041E5)	Biolegend	Cat#356016; RRID:AB_2563974
PE/Dazzle 594 mouse anti-human CD57 (Clone: HNK-1)	Biolegend	Cat#359620; RRID:AB_2564063
PE/Dazzle 594 mouse anti-human CD127 (Clone: A019D5)	Biolegend	Cat#351336; RRID:AB_2563637
PE-Cy5.5 mouse anti-human CD158a,h (Clone: EB6B)	Beckman Coulter	Cat#A66898; RRID:AB_2857330
PE-Cy5.5 mouse anti-human CD158b1/b2,j-PC5.5 (Clone: GL183)	Beckman Coulter	Cat#A66900; RRID:AB_2857331
PE-Cy7 mouse anti-human CD38 (Clone: HB7)	BD Bioscience	Cat#335790; RRID:AB_399969
PE-Cy7 mouse anti-human TNF-α (Clone: MAb11)	BD Bioscience	Cat#557647; RRID:AB_396764
PE-Cy7 mouse anti-human CD56 (Clone: NCAM16.2)	BD Bioscience	Cat#335826; RRID:AB_2857328
PE-Cy7 mouse anti-human Nkp46 (Clone: 9E2)	BD Bioscience	Cat#562101, RRID:AB_10894195
APC mouse anti-human NKp46 (Clone: 9E2)	BD Bioscience	Cat#558051; RRID:AB_398653
Alexa Fluor 700 mouse anti-human Granzyme B (Clone: GB11)	BD Bioscience	Cat#560213; RRID:AB_1645453
Alexa Fluor 700 mouse anti-human CD19 (Clone: HIB19)	Biolegend	Cat#302226; RRID:AB_493751
Alexa Fluor 700 mouse anti-human CD158e1 (Clone: DX9)	Biolegend	Cat#312712; RRID:AB_2130824
APC-Cy7 mouse anti-human CD16 (Clone: 3G8)	Biolegend	Cat#302018; RRID:AB_314218
APC-Cy7 mouse anti-human CD127 (Clone: A019D5)	Biolegend	Cat#351348; RRID:AB_2629572
APC-Cy7 mouse anti-human CD62L (Clone: DREG-56)	Biolegend	Cat#304814; RRID:AB_493582
APC-Cy7 mouse anti-human CD4 (Clone: RPA-T4)	Biolegend	Cat# 300518; RRID:AB_314086
Mouse anti-EBNA2 (Clone: PE2)	Abcam	Cat#ab90543; RRID:AB_2049594
Rat anti-LANA (Clone: LN53)	CliniSciences	Cat#Mob395; RRID:AB_2860565
Rabbit anti-human CD20 (Clone: SP32)	Cell Marque	Cat#120R-16; RRID:AB_2860563
Mouse anti-IRF4/MUM1 (Clone: MUM1p)	CliniSciences	Cat#Mob420; RRID:AB_2861384
Purified anti-human NKp46 (Clone: BAB281)	Beckman Coulter	Custom Purified Bulk Antibody (IMBULK1C)
Goat anti-human IgM-UNLB (polyclonal)	Southern Biotech	Cat# 2020–01 RRID:AB_2795599
Rituximab	Roche Pharma (Schweiz) AG	MabThera
Bacterial and virus strains
EBV B95-8-GFP (EBVwt) produced in HEK293	Delecluse et al., 1998	N/A
EBV B95-8-BZLF1KO-GFP (EBVzko) produced in HEK293	Feederle et al., 2000	N/A
rKSHV.219	Vieira and O’Hearn, 2004; Kati et al., 2015	N/A
Biological samples
Human Fetal Liver samples	Advanced Bioscience Resources	N/A
Chemicals, peptides, and recombinant proteins
Zombie Aqua Fixable Viability Dye	Biolegend	Cat#423102
Zombie NIR Fixable Viability Dye	Biolegend	Cat#423106
LIVE/DEAD Fixable Blue Dead Cell Stain Kit	ThermoFisher Scientific	Cat#L23105
TPA (12-O-Tetradecanoylphorbol 13-acetate)	Sigma-Aldrich	Cat#P1585
Recombinant human IL12 p70	Peprotech	Cat#200-12
Recombinant Human IL-18	Biolegend	Cat#592102
TAPI-1 (ADAM-17 (TACE) and MMP inhibitor)	Tocris	Cat#6162
Critical commercial assays
DNeasy Blood & Tissue Kit	QIAGEN	Cat#69506
MSD U-plex Biomarker Group 1 (Human)	Meso Scale Discovery	Cat#K15067L
MSD V-PLEX Proinflammatroy Panel 1 Human Kit	Meso Scale Discovery	Cat#K15049D
IL-18 Human ProcartaPlex Simplex Kit	ThermoFisher Scientific	Cat#EPX01A-10267-901
Experimental models: Cell lines
Brk.219	Kati et al., 2013	N/A
Experimental models: Organisms/strains
NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG mice)	The Jackson Laboratory	Stock#005557
NOD.Cg-Mcph1 t9(ĥ-a2. DiEnge Prkdcscid Il2rgtm1Wil/SzJ (NSG-A2 mice)	The Jackson Laboratory	Stock#009617
Oligonucleotides
EBV BamHI forward primer: 5′-CTTCTCAGTCCAGCGCGTTT-3′	modified from Berger et al. (2001)	N/A
EBV BamHI reverse primer:5′-CAGTGGTCCCCCTCCCTAGA-3′	modified from Berger et al. (2001)	N/A
EBV BamHI probe:5’-(FAM)-CGTAAGCCAGACAGCAGCCAATTGTCAG-(TAM RA)-3’	modified from Berger et al. (2001)	N/A
KSHV ORF26 forward primer: 5’-GCTCGAATCCAACGGATTTG-3’	modified from Tedeschi et al. (2001)	N/A
KSHV ORF26 reverse primer: 5’-AATAGCGTGCCCCAGTTGC-3	modified from Tedeschi et al. (2001)	N/A
KSHV ORF26 probe: 5’-(FAM)-TTCCCCATGGTCGTGCCTC-(BHQ-1)-3’	modified from Tedeschi et al. (2001)	N/A
Software and algorithms
R (V3.6.3)	R Core Team	https://www.r-project.org
RStudio (v1.2.5033)	RStudio	https://www.rstudio.com
FlowJo (v10.6.2)	FlowJo	https://www.flowjo.com
inform (v2.4.8)	PerkinElmer	N/A
Phenochart (v1.0)	PerkinElmer	N/A
Vectra 3.0	PerkinElmer	N/A
Prism (v8.4.3)	GraphPad Software	https://www.graphpad.com:443/
SPICE - Simplified Presentation of Incredibly Complex Evaluations	Roederer et al., 2011	https://niaid.github.io/spice

Functional assays

Single cell suspensions of liver and spleen tissue were cocultured with K562 or Raji cells at a ratio of 10:1 in RPMI 1640 (GIBCO, Thermo Fisher Scientific) with 10% FCS in a humidified incubator at 37°C and 5% CO₂. Culture medium contained FITC-labeled mouse anti-human CD107a (clone H4A3, BD Bioscience) antibody and monensin was added after 1 hour of coculture. For the ADCC assay, Raji cells were incubated with Rituximab (MabThera, Roche) at a concentration of μg/ml for 10min at 37°C prior to the addition of effector cells. For the stimulation with IL-12 (20ng/ml, Biolegend) and IL-18 (10ng/ml, Biolegend), the culture medium was supplemented with metalloproteinase inhibitor TAPI-1 (5 μg/ml, Tocris), and monensin and Brefeldin A was added after 1 hour of stimulation. After a total of 5 hours, cells were washed, stained, and analyzed on a BD LSR II Fortessa or BD FACSymphony A5 flow cytometers as described above.

Histological staining

Spleen and liver sections were fixed in 4% formalin and paraffin-embedded (FFPE). Immunohistochemical single staining of EBNA2 (clone PE2, Abcam) and LANA (clone LN53, Clinisciences) was carried out on a Leica BOND-III automated immunohistochemistry system using diaminobenzidin (DAB) as chromogen (Zytomed Systems, Berlin, Germany). In situ-hybridization for the detection of EBER was performed as described previously (Meyer et al., 2011) with DAP as substrate. Multiplex immunofluorescence (IF) staining was performed on FFPE tissue using Opal dyes and Spectral DAPI (FP1490) from PerkinElmer. Opal dyes 520, 540, 620 and 690 were used to detect EBNA2 (clone PE2, Abcam), CD20 (clone SP32, Cell Marque), LANA (clone LN53, CliniSciences) and IRF4/Mum1 (clone: MUM1p, CliniSciences), respectively. Single stains were used to compensate overlapping spectras. Staining was quantified on a Vectra3 automated quantitative pathology imaging system using Vectra, Phenochart (v1.0) and InForm (v2.4.8) software (all from PerkinElmer), as previously described (Murer et al., 2018).

Serum cytokine quantification

Serum of huNSG mice was collected upon terminal cardiac puncture using BD Mictrotainer tubes and frozen at −20°C until use. Samples were assayed for cytokines using the Meso Scale Discovery (Meso Scale Diagnostics, Rockville, MD) U-PLEX (Biomarker Group 1 Human, K15067L) and V-PLEX (Proinflammatory Panel 1 Human Kit, K15049D) platform following the manufacturer’s recommendations. Plates were read on a Meso QuickPlex SQ120 and analyzed with Discovery Workbench 4.0.12, both from Meso Scale Discovery. Samples and calibrator dilutions for the standard curves were run in duplicates. Heatmaps were generated in R using the superheat package (Barter and Yu, 2018) after empirical percentile transformation of cytokine levels (heatmaply package; Galili et al., 2018).

Plasma from Kenyan children samples was isolated from whole blood and kept frozen until use. IL-18 concentration in plasma of Kenyan children was measured using Human IL-18 Simplex ProcartaPlex from eBiosciences (cat#EPX01A-10267-901, lot#144494103, ULOQ/LLOQ of 49500/12 pg/ml). The assay was performed following the manufacturer’s protocol and run on a Bio-Plex 200 (Biorad) platform using the Bioplex Manager Software (Biorad, version 6.1). The concentration of the samples was calculated by plotting the expected concentration of the standards against the MFI generated by each standard using a5PL algorithm.

QUANTIFICATION AND STATISTICAL ANALYSIS

Survival curves were analyzed by log-rank test. Contingency tables of categorical data were examined by two-tailed Fisher’s exact test. Correlation was assessed by Spearman’s rank test. The D’Agostino & Pearson omnibus test was used to test the normal distribution of data. Unpaired non-parametric data were analyzed by Mann-Whitney U test or by Dunn’s multiple comparison test following Kruskall Wallis test when comparing more than two groups. Paired non-parametric data were analyzed by Wilcoxon signed-rank test or by multiple comparison tests following Friedmann’s test when comparing more than two groups. Paired parametric data were analyzed by repeated-measures ANOVA or by fitting a mixed model (REML) in case of missing values, followed by Fisher’s LSD post hoc test. P values from follow-up multiple comparison tests are FDR-adjusted. A p or FDR-adjusted q value of < 0.05 was considered statistically significant. All statistics were computed with GraphPad Prism 8 (GraphPad Software, San Diego, CA) and details can be found in figures or in corresponding legends.

Highlights.

• CD56⁻ CD16⁺ NK cells accumulate upon KSHV/EBV dual infection in humanized mice

• KSHV titers correlate with CD56⁻ CD16⁺ NK cell frequency in vivo

• CD56⁻ CD16⁺ NK cells display impaired effector functions