CD38 contributes to human natural killer cell responses through a role in 1 immune synapse formation

Natural killer (NK) cells use a diverse array of activating and inhibitory surface receptors to detect threats and provide an early line of defense against viral infections and cancer. Here, we demonstrate that the cell surface protein CD38 is a key human NK cell functional receptor through a role in immune synapse formation. CD38 expression marks a mature subset of human NK cells with a high functional capacity. NK cells expressing high levels of CD38 display enhanced killing and IFN-γ secretion in response to influenza virus-infected and tumor 25 cells. Inhibition of CD38 enzymatic activity does not influence NK cell function, but blockade of CD38 and its 26 ligand CD31 abrogates killing and IFN-γ expression in response to influenza-infected cells. Blockade of CD38 on NK cells similarly inhibits killing of tumor cells. CD38 localizes and accumulates at the immune synapse 28 between NK cells and their targets, and blocking CD38 severely abrogates the ability of NK cells to form 29 conjugates and immune synapses with target cells. Thus, CD38 plays a critical role in NK cell immune synapse 30 formation. These findings open new avenues in immunotherapeutic development for cancer and infection by revealing a critical role for CD38 in NK cell function.


8
Cell mixtures were run on a three-laser MACSQuant® Analyser (Miltenyi). Analysis and compensation were 5 9 performed using FlowJo flow-cytometric analysis software, version 9.9.4 (Tree Star). Percent of conjugated NK 6 0 cells to their targets were calculated using this formula: Confocal microscopy 6 5 Isolated NK cells were stained using CellTrace Violet dye (Thermofischer) for 20 min at RT then washed twice 6 6 in PBS. Isolated NK and mock-or H1N1-infected monocytes were incubated on a Poly-L-Lysine pre-coated 8 6 7 well µ-slide (Ibidi) for 2 hours. Cells were then washed in PBS-FBS 2%, fixed in PFA 4% for 15 min and 6 8 washed twice in PBS-FBS 2%, then stained with mouse anti-CD38 and/or rabbit anti-LFA-1 antibody for 30 6 9 min at RT, then washed twice in PBS-FBS 2%. Secondary staining was performed using a goat anti-mouse 7 0 AlexaFluor594 or a goat anti-rabbit AlexaFuor488 antibody for 30 min at RT. After washing the cells twice in 7 1 PBS-FBS 2%, cell mount media (Ibidi) was added for microscopy. Images were acquired using LSM880 Meta 7 2 (Zeiss) laser scanning confocal microscope equipped with a 63× (NA 1.4) DIC oil objective. All counting of 7 3 synapses was performed in a blinded fashion. Given that CD38 is highly expressed in NK cells (Fig. S1A), is a marker of lymphocyte activation, and is part 8 7 of a highly accurate diagnostic of influenza infection, we hypothesized that it might be induced on NK cells 8 8 during acute influenza infection. Indeed, acutely influenza-infected individuals displayed a very robust increase 8 9 in CD38 expression on NK cells, which returned to baseline levels during convalescence (Fig. 1A, Table S1, 9 0 Fig. S1B). As all of the healthy controls happened to be male, this raised the possibility that there may be sex-9 1 related differences in CD38 expression contributing to the differences observed. However, we observed no 9 2 significant differences between men and women in NK cell expression of CD38, making this an unlikely 9 3 possibility (Fig. S1C). 9 4 9 5 To determine whether CD38 expression is similarly modulated on NK cells encountering infected cells ex vivo, 9 6 we exposed NK cells to autologous influenza-infected monocytes as previously described (He et al., 2004;Kay 9 7 et al., 2014;Kronstad et al., 2018). NK cells themselves are not infected by influenza in this system(Kronstad et 9 8 al., 2018). We observed a significant increase in NK cell CD38 expression upon their encounter with monocytes 9 9 infected with the pandemic A/H1N1 influenza strain (H1N1) compared to mock-infected conditions (Fig. 1B). 0 0 We also observed significantly increased expression of CD107a, a marker of degranulation, and IFN-γ ( in CD38 high vs. CD38 low NK cells ( Fig. 1E and F). Thus, CD38 expression is induced upon encounter with 0 5 influenza-infected cells and its expression is associated with enhanced NK cell antiviral function. 0 6 0 7 Profiling CD38 low and CD38 high NK cells. 0 8 To understand how CD38 contributes to enhanced NK cell function, we explored the phenotype of CD38 0 9 expressing NK cells using mass cytometry (Fig. S1, Table S4). We used a Generalized Linear Model (GLM) 1 0 with bootstrap resampling to identify markers predictive of CD38 high vs. CD38 low NK cells in a discovery NKp30, NKp44 and NKp46) and inhibitory (KIR2DL1, KIR2DL3, KIR2DL5, KIR3DL2 and LILRB1) 1 4 receptors, while CD38 low NK cells were more likely to express the activating receptor NKG2D, and the 1 5 inhibitory receptors NKG2A and KIR2DL4 ( Fig. 2A). A similar analysis of 21 healthy women in a validation 1 6 cohort using a different antibody panel confirmed the increased expression of NKp30 and NKp46, but not 1 7 NKp44, in CD38 high NK cells compared to CD38 low NK cells (Fig. 2B, Table S4, Table S5). We also confirmed 1 8 the decreased expression of NKG2A, but not NKG2D, among CD38 high NK cells in this validation cohort. 1 9 Expression of perforin, which was only evaluated in the validation cohort, was significantly predictive of 2 0 CD38 high NK cells. Manual gating confirmed these results for both cohorts (Fig. S3, S4, S5A and S5B).

1
Overall, the increased expression of several KIRs, NKp30, NKp46, CD11b, and perforin, along with decreased 2 2 expression of NKG2A, suggest that CD38 high NK cells are a mature subset with high cytotoxic potential.
CD38 plays a direct role in enhanced NK cell function to influenza-infected cells.

5
To determine whether CD38-expressing NK cells have any enhanced functions, we sorted CD38 low and 2 6 CD38 high NK cells and assessed a number of immune response indicators (Fig. S6A). We found that compared 2 7 to CD38 low cells, CD38 high NK cells had a 3.5-fold greater expression of CD107a, 7.2-fold greater IFN-γ 2 8 expression, and 2.2-fold greater ability to induce the death of influenza infected monocytes (Fig. 3A, B and C).

9
To directly test the role of CD38 in this enhanced function, we assessed the potential contributions of CD38-3 0 CD31 mediated adhesion using the previously reported agonistic anti-CD38 antibody IB4 and blocking 3 1 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/349084 doi: bioRxiv preprint antibodies to both CD38 and its ligand, CD31, which is expressed on monocytes (Fig. S6B). Given the critical 3 2 role of LFA-1 in immune synapse formation (Orange, 2008), blocking antibodies to LFA-1 were also used as a 3 3 positive control. In a distinction from the setting of IL-2 activated NK cells, the agonistic IB4 antibody did not 3 4 have a significant effect on NK cell function in response to influenza-infected cells (Fig. S6C). However, 3 5 blocking CD38 and its ligand CD31 led to a marked reduction in NK cell function in response to influenza-3 6 infected monocytes, with 2.4-fold reduction in CD107a expression, 2.2-fold reduction in IFN-γ production, and 3 7 1.7-fold reduction in killing of infected monocytes with CD38 blocking (Fig. 3D, E and F). Blocking CD31 and 3 8 LFA-1 also significantly abrogated NK cell expression of CD107a and IFN-γ. In light of concerns for blocking 3 9 antibodies driving antibody-dependent cellular cytotoxicity via the FcRγIII receptor (CD16) on NK cells, we 4 0 repeated the experiments with Fabs derived from CD38 and CD31 blocking antibodies. Both Fabs significantly 4 1 decreased the NK cell CD107a and IFN-γ response to influenza-infected monocytes, just as the parent blocking 4 2 antibodies did ( Fig. S6D and E). We next explored whether the enzymatic activity of CD38 might play a role in 4 3 NK cell function to influenza-infected cells. The CD38 enzymatic inhibitor, kuromanin, did not significantly and G). Thus, CD38 appears to contribute to NK cell responses against infected cells through its role in binding 4 6 to CD31. This led us to investigate whether CD38-CD31 interactions play a role in the formation of the immune synapse 5 2 between NK cells and infected cells. Blocking antibodies to both CD38 and LFA-1, which was used as a 5 3 positive control, led to a significant 1.89-and 1.97-fold reduction in the formation of NK cell-monocyte 5 4 conjugates compared to untreated cells or isotype control antibody-treated cells, respectively ( Fig. 4A and Fig.  5  5 S7A), suggesting that CD38 is important for adhesion between NK cells and their targets. Using confocal 5 6 microscopy, we found that CD38 localizes to the immune synapse at the contact zone between NK cells and 5 7 H1N1-infected monocytes, which was identified based on co-staining with the adhesion molecule LFA-1 (Fig.  5  8 4B and C). Minimal accumulation of CD38 was observed when NK cells were in contact with mock-infected 5 9 monocytes. In the presence of blocking CD38 or CD31 antibodies, the formation of the immune synapse was 6 0 strongly and significantly inhibited compared to cells incubated with an isotype control antibody (2.7-and 2.0-6 1 fold reduction, respectively, Fig. 4D and E). Similarly, immune synapse formation was rarely observed between 6 2 H1N1-infected cells and CD38 low NK cells. Together, these data indicate that CD38 contributes to the 6 3 establishment of the immune synapse between NK cells and influenza-infected cells, which is critical for NK 6 4 cell immune responses. 6 5 6 6 CD38 contributes to immune synapse formation between NK cells and cancer cells.
To explore whether the role of CD38 extends beyond the response to influenza-infected cells, we examined NK 6 8 cell responses to K562 tumor cells (Lisovsky et al., 2015). NK cells upregulated CD38 in response to K562 6 9 cells (Fig. S8A). Sorted CD38 high NK cells were significantly better than CD38 low NK cells at killing tumor 7 0 cells ( Fig. 5A and B) and secreting IFN-γ ( Fig 5C). Blocking CD38 significantly diminished CD107a 7 1 expression on NK cells and inhibited tumor cell killing ( Fig. 5D and E), but not IFN-γ production (Fig. 5F). 7 2 Blocking CD31 had no significant effect on CD107a expression, K562 cell death, or IFN-γ production, 7 3 suggesting that CD38 may bind an alternate ligand on these tumor cells. Blocking CD38 or LFA-1 significantly 7 4 diminished conjugate formation with tumor cells (Fig. 5G and Fig. S8B). Further, sorted CD38 high NK cells 7 5 have a significantly greater potential to form conjugates with K562 cells than do CD38 low NK cells (Fig. 5H  7  6 and Fig. S8C). As in the synapse with infected cells, CD38 polarizes to the immune synapses between NK cells 7 7 and K562 tumor cells ( Fig. 5I and Fig S8D). These results indicate that CD38 plays a crucial role in the 7 8 establishment NK cell cytotoxic function and immune synapse formation with cancer cells. 7 9 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/349084 doi: bioRxiv preprint Here, using mass cytometry and functional assessments, we demonstrate that CD38 expression marks a mature 8 7

Discussion
subset of NK cells with high cytolytic and cytokine producing capacity. Further, our study reveals that CD38 8 8 plays a critical role in NK cell function against both viral-infected and tumor cells by enhancing immune 8 forward in our understanding of the precise mechanisms. CD2, LFA-1 and CD38 may also physically interact 2 9 and serve to stabilize or enhance NK cell adhesion to their targets. Defining the nature of such interactions will 3 0 be an important area of further study. 3 1 3 2 During the effector phase, the immune synapse plays an essential role in NK cell activation. This activation is 3 3 thought to be partly mediated by LFA-1 intracellular signaling (Barber et al., 2004;Orange, 2008). CD38 has a 3 4 short cytoplasmic domain without any known signaling motifs, but can potentiate signaling through the CD3-3 5 TCR complex in T cells, MHC class II in monocytes, CD19 in B cells (Lund et al., 1996;Zilber et al., 2005;3 6 Zubiaur the identification of a new role for CD38 in immune synapse formation and optimal cytotoxic activity has 4 8 significant implications for these therapies. It will be important to understand the factors that control CD38 4 9 expression in vivo to assure its retention on engineered cells. For instance, we find that both IL-2 and IL-15, two 5 0 important cytokines for NK cell homeostasis, promote increased CD38 expression in vitro (data not shown).

1
These data also raise interesting questions about the CD38-targeting agents that are currently in clinical use, 5 2 primarily for multiple myeloma and chronic lymphocytic leukemia (Damle et al., 1999;Lokhorst et al., 2015).

3
Clinical treatment with daratumumab, a specific human CD38 binding antibody, improves patient A limitation of our study is the fact that our mass cytometry panels differed between the two cohorts and that 6 1 this phenotypic data was exclusively on female subjects. To ameliorate this concern, all functional experiments 6 2 were performed with samples from healthy blood bank donors of both sexes. A second limitation is that we 6 3 used blocking antibodies or Fab and sorted populations rather than knockouts to establish CD38 function. While 6 4 CRISPR-Cas9-mediated editing of NK cells is feasible(Phan et al., 2016; Somanchi and Lee, 2016), the 6 5 expansion protocol required leaves NK cells unable to respond to influenza-infected cells (not shown). Thus, 6 6 even if we were able to knockout CD38, we could not appropriately assess its impact on NK cell function in this 6 7 physiologic antiviral response. 6 8 6 9 In summary, our goal was to better understand the role of CD38 in NK cell biology and activity in the context 7 0 of influenza A virus infection and cancer. We find that, at least in vitro, CD38 expression is crucial in both 7 1 settings. We also demonstrate a novel role for CD38 in the establishment of immune synapses that are 7 2 characteristic of efficient killing by NK cells. This will be useful for potential NK cell engineering and other 7 3 therapeutic applications. 7 4 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/349084 doi: bioRxiv preprint 1 1

3
Antigen-induced clustering of surface CD38 and recruitment of intracellular CD38 to the immunologic synapse.

7
A GLM with bootstrap resampling was used to identify markers predictive of CD38 low vs CD38 high NK cells 5 8 within the discovery cohort (A) and validation cohort (B). The markers studied are listed vertically and the x-5 The copyright holder for this preprint (which was not peer-reviewed) is the author/fund . https://doi.org/10.1101/349084 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the author/fund . https://doi.org/10.1101/349084 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the author/fund . https://doi.org/10.1101/349084 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the author/fund . https://doi.org/10.1101/349084 doi: bioRxiv preprint