CD38 is a key regulator of enhanced NK cell immune responses during pregnancy through its role in immune synapse formation

Once sentence summary CD38 is responsible for the enhanced immune responses of NK cells to influenza virus infection during pregnancy through immune synapse formation. Abstract Pregnant women are particularly susceptible to complications of influenza A virus infection, which may result from pregnancy-induced changes in the function of immune cells, including natural killer (NK) cells. To decipher mechanisms driving enhanced NK cell activity during pregnancy, we profiled NK cells from pregnant and non-pregnant women, which showed significantly increased CD38 expression during pregnancy. CD38 expression defines a phenotypically distinct and mature subset of NK cells that display increased ability to secrete IFN-γ and to kill influenza-infected and tumor cells. This enhanced function is based on the ability of CD38 to promote the formation of the NK cell immune synapse. Thus, increased CD38 expression directly promotes enhanced NK cell responses during pregnancy through its role in immune synapse formation. These findings open new avenues in immunotherapeutic development for cancer and viruses by revealing a critical role for CD38 in the formation of the NK cell immune synapse.

pregnancy. CD38 expression defines a phenotypically distinct and mature subset of NK cells that display 26 increased ability to secrete IFN-g and to kill influenza-infected and tumor cells. This enhanced function is 27 based on the ability of CD38 to promote the formation of the NK cell immune synapse. Thus, increased 28 CD38 expression directly promotes enhanced NK cell responses during pregnancy through its role in 29 immune synapse formation. These findings open new avenues in immunotherapeutic development for 30 cancer and viruses by revealing a critical role for CD38 in the formation of the NK cell immune synapse. 31

Introduction 33
Pregnant women are at high-risk of complications from seasonal and pandemic influenza infections (1,2). 34 During the 2009 H1N1 influenza virus pandemic, pregnant women in the United States suffered a 35 disproportionately high mortality rate, accounting for 5% of deaths while representing only 1% of the total 36 population (3). The majority of pregnant women who died of influenza-related illness during the pandemic 37 were infected in the second and third trimesters of pregnancy (4). Influenza infection during second or third 38 trimester of pregnancy is also associated with significant increases in miscarriages, stillbirths, and early 39 neonatal diseases and death (1,5). However, the mechanisms behind this susceptibility to influenza 40 infection during pregnancy are still poorly understood. 41 42 During pregnancy, the immune system has to finely balance its activity in order to tolerate semi-allogenic 43 fetal antigens, while maintaining the ability to fight microbial challenges (6-9). These immune alterations 44 may be at least partially responsible for the increased susceptibility of pregnant women to influenza virus 45 including NKp46, NKp30 and NKG2D, which together define their degree of maturation and 51 responsiveness to stimuli (19,20). The expression of these receptors enables NK cells to sense altered cells, 52 including virus-infected and cancerous cells (21)(22)(23). In response to such threats, NK cells can produce 53 cytokines, such as IFN-g, which limits viral replication and tumor proliferation, and kill cells via release of 54 cytolytic molecules or through engagement of death receptors. NK cell activation must be tightly regulated 55 to limit tissue damage at the site of infection. The mechanisms by which NK cell activity is altered during 56 pregnancy are unknown but are particularly critical to balance protection from infection with the avoidance 57 of a hyperinflammatory reaction that could cause irreversible injuries to the mother and the fetus. 58 59 The goal of our study was to identify pregnancy-mediated changes in NK cells and their impact on the 60 immune response to influenza infection. Our prior study reported that NK cell responses to influenza virus-61 infected cells were enhanced during pregnancy (12). This finding may at least in part reflect the fact that 62 virus infections were performed in bulk peripheral blood mononuclear cell (PBMC) cultures, where 63 enhanced cytokine responses from monocytes and dendritic cells could have contributed to dramatic NK 64 cell activation (13,21). In fact, NK cells from pregnant women stimulated directly with phorbol-myristate 65 acetate and ionomycin in our experiments, or with IL-12/IL-15 by Kraus et al. (15), have suppressed 66 responses compared to those of non-pregnant women. In this study, we aim to resolve these contradictory 67 findings by determining the mechanisms behind changes in cell-intrinsic NK cell function during 68 pregnancy. To do so, we used mass cytometry paired with logistic regression to predict pregnancy-related 69 changes in NK cells between pregnant and non-pregnant women in two independent cohorts. We performed 70 functional assays to define the role of specific NK cell receptors in the responsiveness to influenza-infected 71 cells, uncovering a surprising new role for the cell surface molecule CD38 in the formation of the immune 72 synapse that is required for NK cell cytotoxic activity. We then explored whether this phenomenon extended 73 beyond the setting of influenza-infected cells and discovered that CD38 also promotes immune synapse 74 formation between NK cells and cancer cells, potentially presenting new therapeutic targets in cancer and 75 infectious diseases. 76

NK cell immune response to influenza virus and cancer cells during pregnancy. 78
To investigate how pregnancy alters NK cell phenotype and function, we recruited two cohorts of pregnant 79 and non-pregnant (control) women in subsequent years (Tables S1 and S2 for cohort 1 and 2 demographics). 80 Pregnant women were enrolled in their second or third trimesters, and blood samples were collected at 81 enrollment and 6 weeks postpartum. To investigate intrinsic NK cell function during pregnancy, NK cells 82 and autologous monocytes were sorted from the PBMCs of controls (n=10), pregnant women (n=10), and 83 postpartum women (n=10) from cohort 1 (Fig. 1A, experimental workflow). Monocytes were infected with 84 2009 pandemic H1N1 influenza virus strain and then co-cultured with the autologous NK cells before 85 functional assessment by flow cytometry. We observed that the frequency of NK cells expressing CD107a 86 as a marker of cytolytic activity (Fig. 1B) and IFN-g production ( Fig. 1C) was significantly greater in 87 pregnant women than in controls or in postpartum women. These data confirm our earlier findings that 88 influenza-specific NK cell responses are enhanced during pregnancy (12). However, monocytes 89 demonstrate enhanced anti-influenza responses during pregnancy, which could activate NK cells through 90 inflammatory cytokine production (13). We hypothesized that if NK cell function was intrinsically elevated 91 during pregnancy, we should observe enhanced anti-tumor responses as well. We therefore exposed sorted 92 NK cells from controls and pregnant women to the K562 tumor cell line (Fig. 1A), which represents a 93 homogenous, identical target for NK cells from controls and pregnant women. NK cells from pregnant 94 women demonstrated significantly greater CD107a expression and tumor cell killing than NK cells from 95 non-pregnant women (Fig. 1, D and E). Interestingly, though tumor killing activity was enhanced during 96 pregnancy, there were no significant differences in the frequency of IFN-g-producing NK cells (Fig. 1F), 97 possibly reflective of the lack of inflammatory cytokine production by these tumor targets. These data 98 indicate that NK cells have an intrinsically enhanced ability to kill both infected and tumor targets during 99 pregnancy. 100

101
Deep profiling of NK cell phenotype from healthy pregnant and non-pregnant women 102 To understand potential drivers of this enhanced NK cell function during pregnancy, we next profiled the 103 expression patterns of inhibitory and activating surface receptors on NK cells in control non-pregnant 104 women, pregnant women, and postpartum women. PBMCs from women in both cohorts were evaluated by 105 mass cytometry as outlined in Fig. 2A and Tables S3 and S4 (antibody panels for cohort 1 and 2, 106 respectively). NK cells were identified as CD3 -CD19 -CD20 -CD14 -CD56 +/-CD16 +/cells (Fig. S1A). The 107 frequency of NK cells did not significantly differ between pregnant and control women, nor in pregnant vs. 108 postpartum women in either cohort (Fig. S1, B and C). To identify NK cell markers predictive of pregnancy, 109 we used a generalized linear model (GLM) with bootstrap resampling to account for correlations between 110 cells and inter-individual variability ( Figure 2A, and Materials and Methods). This method was selected 111 because markers expressed on cells from the same subject are usually more highly correlated than marker 112 expression patterns between subjects. Thus, accounting for correlated data reduces the number of false 113

positives. 114
In both cohorts, the expression of CD38 was significantly predictive of pregnant vs. non-pregnant 115 women (Fig. 2, B and C). CD38 also strongly predicted pregnancy in a paired comparison within women 116 between their pregnant and post-partum time points (Fig. 2, D and E). To confirm these results, CD38 117 expression on NK cells was assessed by manual gating, revealing that NK cells from pregnant women 118 express significantly more CD38 than NK cells from non-pregnant or postpartum women (Fig. S2, A to F). 119 Several other markers, including PD-1, CD27, CD94, LILRB1, NKp46, NKp44, NKG2A, and CD244, 120 were predictive of pregnancy in one of the cohorts or analyses (Fig. 2, B to E). Manual gating analyses 121 similarly confirmed enhanced expression of several of these markers within individual cohorts and 122 comparisons, yet none of the markers was consistently predictive of pregnancy across cohorts or analyses 123 ( Fig. S3 to S7). As CD38 was the only marker consistently predictive of pregnancy across both cohorts and 124 all analyses, we elected to explore how its enhanced expression during pregnancy might alter NK cell 125 function. 126 127 CD38 expression on NK cells correlates with the magnitude of the influenza-specific immune response 128 We assessed CD38 expression by conventional flow cytometry on NK cells during their response to 129 influenza-infected cells as outlined in Fig. 1A. Consistent with our mass cytometry results, the frequency 130 of CD38-expressing NK cells as well as the intensity of CD38 expression on NK cells was significantly 131 greater in pregnant women compared to controls or postpartum women (Fig. 3, A and B). Further, CD38 132 expression was significantly correlated with the magnitude of the NK cell response as measured by CD107a 133 expression (Fig. 3C) and IFN-g production (Fig. 3D). This raises the possibility that the increased responses 134 in pregnant women are driven primarily by their enhanced CD38 expression. We therefore evaluated the 135 function of CD38vs. CD38 + NK cells. Regardless of pregnancy status, CD38-expressing NK cells 136 displayed a higher CD107a and IFN-g expression compared to CD38 -NK cells (Fig. 3, E and F). These data 137 are consistent with the idea that expansion of a highly functional subset of NK cell expressing CD38 is 138 responsible for the enhanced influenza-specific NK cell responses during pregnancy. 139 To further confirm that CD38 expression is associated with enhanced NK cell responses to influenza 140 regardless of pregnancy status, we used NK cells from healthy blood bank donors to compare the ability of 141 CD38versus total NK cells to respond to pH1N1-infected monocytes. We elected to compare CD38 -NK 142 cells to total NK cells (of which 45-95% express CD38, Fig. S2) rather than to sort on CD38 because its 143 ligation could alter NK cell function. Total NK cells displayed significantly greater CD107a and IFN-g 144 responses to autologous influenza-infected monocytes than did CD38 -NK cells (Fig. 3

, G and H). 145
Moreover, the frequency of dead monocytes was significantly increased after co-culture with total NK cells 146 compared to CD38 -NK cells, suggesting that CD38 + NK cells are enhanced in killing activity (Fig. 3I). 147 These data confirm that CD38 + NK cells have increased responses to influenza-infected cells.  To better understand the mechanisms driving enhanced function of CD38-expressing NK cells, we 151 examined the phenotype of CD38 + NK cells. We used GLM with bootstrap resampling to evaluate markers 152 predictive of CD38 + vs. CD38 -NK cells (Fig. 4A, for cohort 2 and fig. S8A, for cohort 1), which differed 153 in the specific NK cell markers examined ( Fig. 2A). As CD38 expression was associated with enhanced 154 responsiveness in all subjects regardless of pregnancy status, we examined NK cells from both pregnant 155 and control women together. We observed that CD38 expression marked a unique phenotype of NK cells 156 in cohort 2 (Fig. 4A). Specifically, several activating (CD244 (2B4), CD11b, CD57, NKp30 and NKp46) 157 and inhibitory (KIR2DL1, KIR2DL2/L3/S2, KIR3DL1, KIR3DL2 and LILRB1) receptors were 158 significantly predictive of CD38 + NK cells. CD38 -NK cells were more likely to express CD94, CD27, and 159 KIR2DL1. A similar analysis of cohort 1 confirmed the increased expression of CD57, NKp30 and NKp46 160 in CD38 + NK cells compared to CD38 -NK cells (Fig. S8A). Expression of perforin, which was not included 161 in the panel to evaluate cohort 2, was significantly increased on CD38 + NK cells (Fig. S8A). Manual gating 162 to examine the frequency of NK cells expressing specific receptors, as well as their mean signal intensity, 163 confirmed these results (Fig. 4, B and C). Similar results were obtained when the control and pregnant 164 women were evaluated independently (Fig. S9, A and B). Overall, the increased expression of several KIRs, 165 NKp30, NKp46, CD57, CD11b, and perforin suggest that CD38 + NK cells are a mature subset with high 166 cytotoxic potential. 167

CD38 plays a direct role in NK cell immune response to influenza infection. 169
The mature phenotype of CD38-expressing NK cells could explain their enhanced responsiveness if CD38 170 marks cells with greater functional potential. Yet it is also possible that CD38 plays a direct role in the NK 171 cell immune response to influenza-infected cells. CD38 has two known functions, neither of which have 172 been studied in antiviral or antitumoral NK cell function. First, it is an extracellular ectoenzyme that can 173 drive intracellular calcium flux through i) catalyzing the synthesis of cyclic Adenosyl-di-phosphate ribose 174 from Nicotinamide Adenine Dinucleotide + or ii) catalyzing the hydrolysis of cyclic Adenosyl-di-phosphate 175 ribose into Adenosyl-di-phosphate ribose (22, 23). Second, it is an adhesion molecule that binds to CD31 176 and likely other ligands (24, 25). We therefore used an inhibitor of CD38 enzymatic activity, kuromanin, 177 to assess the role of CD38 enzymatic activity in the NK cell response to influenza-infected monocytes. We 178 observed no significant differences in CD107a or IFN-g responses in the presence or absence of this 179 inhibitor at concentrations known to inhibit CD38 function in T cells (26) (Fig. S10, A and B). Thus, the 180 enzymatic activity of CD38 does not appear to play a role in altering influenza-specific NK cell responses. 181 As monocytes express CD31 (Fig. S10C), we used CD38 and CD31 blocking antibodies to assess 182 the role of CD38-CD31 interactions in the immune response to influenza. The addition of CD38 blocking 183 antibodies significantly inhibited NK cell CD107a and IFN-g responses (Fig. 5, A and B). Further, blocking 184 CD38 diminished the frequency of dead monocytes, suggesting that NK cell killing is diminished (Fig. 5C). 185 Blocking the CD38 ligand, CD31, also significantly abrogated NK cell CD107a and IFN-g responses 186 compared to the incubation with an isotype control antibody. These data indicate that CD38, in addition to 187 marking a mature subset of NK cells, plays a direct role in NK cell responses to influenza, likely by the 188 binding to its ligand, CD31. 189

CD38 is crucial to the establishment of immune synapse with influenza-infected cells. 191
We next investigated whether CD38-CD31 interactions might play a role in the formation of the immune 192 synapse between the NK cell and the infected cell. We therefore examined whether CD38-CD31 193 interactions contributed to the ability of NK cells to make conjugates with influenza-infected monocytes. 194 Conjugation between fluorophore-labelled NK cells and monocytes were assessed by flow cytometry (Fig. 195 5D and fig. S11). In presence of blocking antibodies to both CD38 and LFA-1 (which was used as a positive 196 control), we observed a significant reduction in the formation of NK cell-monocyte conjugates compared 197 to untreated cells or isotype control antibody-treated cells. This suggests that CD38 plays a role in immune 198 synapse formation. To confirm this, using confocal microscopy, we observed that CD38 is co-localized 199 with the adhesion molecule LFA-1 in the contact zone between NK cells and H1N1-infected monocytes 200

CD38 contributes to immune synapse formation between NK cells and cancer cells. 211
To explore whether the role of CD38 in NK cell responses extends beyond the response to influenza-212 infected cells, we examined NK cell responses to K562 tumor cells. We exposed total or CD38 -NK cells 213 from healthy donors to K562 cancer cells in the presence of CD38 or CD31 blocking antibodies. Total NK 214 cells had significantly greater CD107a responses than did CD38 -NK cells and resulted in higher levels of 215 tumor cell killing (Fig. 6, A and B). Blocking CD38 significantly diminished NK cell CD107a expression 216 and tumor cell killing (Fig. 6, A and B). The frequency of IFN-g + cells did not significantly differ between 217 CD38and total NK cells, and blocking CD38 and CD31 did not significantly alter the frequency of IFN-g + 218 NK cells responding to tumor targets (Fig. 6C). Blocking CD31 had no significant effect on CD107a 219 expression, K562 cell death, or IFN-g production, suggesting that an alternate ligand may be used. Thus, 220 these data indicate that CD38 also plays a critical role in NK cell cytolytic responses to cancer cells. To 221 demonstrate the direct role of CD38 in NK cell anti-tumor responses, we assessed NK cell-K562 cell 222 conjugate formation, and found that blocking CD38 or LFA-1 (positive control) significantly diminished 223 conjugate formation compared to isotype control or without antibody (Fig. 6D). Finally, we visualized 224 CD38 in the formation of immune synapses between NK cells and K562 tumor cells. As depicted in Fig. 6, 225 E and F, CD38 polarizes at the contact zone between NK cells and K562 cells. Together, our data indicate 226 that CD38 plays a crucial role in the establishment of the immune synapse and in NK cell cytotoxic function 227 towards cancer cells. 228

Discussion 229
During pregnancy, the maternal immune system is engaged in a fine balance: tolerance is required to 230 preserve the fetus while defenses must be maintained to protect mother and baby from microbial challenges. 231 NK cells play a critical role in this balance as their job is to patrol the body for 'altered self'(27). NK cell 232 activity had been thought to be suppressed during pregnancy to protect the fetus (28) to better understand the potential contributions of NK cells to influenza pathogenesis during pregnancy, we 264 sought to define exactly how pregnancy altered human NK cells. 265

266
We were surprised to discover that the most dramatic and consistent difference between NK cells of 267 pregnant and non-pregnant women was an increase in CD38 expression during pregnancy. While CD38 is 268 most commonly viewed as an activation marker on T cells, it is more highly expressed on NK cells and has 269 several important functions. First, CD38 confers lymphocytes with the ability to adhere to endothelial cells 270 through its binding to CD31, a necessary step in extravasation. CD38 also functions as an ectoenzyme, 271 converting extracellular NAD + to cADPR through its cyclase activity or cADPR to Adenosyl-di-phosphate 272 ribose through its hydrolase activity (35). These molecules in turn can diffuse into the cell and promote its 273 activation by driving intracellular calcium increase, phosphorylation of signaling molecules, production of 274 cytokines, and vesicular transport (35). The diverse roles and functions raised the question of whether 275 increased CD38 expression during pregnancy is merely a marker of a more 'activated' state or whether it 276 was playing a direct role in modulating NK cell function. Thus, we first explored whether CD38 expression 277 marked a unique subset of NK cells. We found that expression of CD38 marks a population of NK cells 278 with a differential expression of KIRs, CD11b, CD57, perforin and CD94 compared to CD38 -NK cells, 279 indicating a higher degree of cell maturation (36). In particular, the high expression of one or several KIRs 280 on CD38 + NK cells is an essential feature characterizing NK cell maturation and terminal differentiation 281 (37, 38). This suggests that CD38 + NK cells are functionally competent, with high cytotoxic abilities, but 282 also likely self-tolerant since they display at least one inhibitory receptor for Major Histocompatibility 283 Complex. The high degree of co-expression between KIR and CD38 could reflect a need to assure tolerance 284 towards fetal semi-allogenic antigens, while maintaining a high degree of cytotoxicity towards virus-285 infected and transformed cells, as demonstrated here. 286

287
Our demonstration here that CD38 contributes to the formation of immune synapse between NK cells and 288 their targets represents a novel role for this molecule. While much is known about its enzymatic activities 289 and adhesion functions through interaction with CD31, CD38 has not been extensively studied in terms of 290 the biology of immune cells, especially cytotoxic ones. We show here that CD38-CD31 interactions are 291 necessary to establish the NK immune synapse and for its cytotoxic activity against influenza-infected target 292 cells. Importantly, this function for CD38 was not restricted to NK cells role in responding to influenza-293 infected cells, as we also show that blocking CD38 diminishes the ability to kill tumor targets. Thus, CD38 294 has a previously unrecognized, but critical, role in facilitating cytolytic activity against both infected cells 295 and tumor targets. Interestingly, blocking CD38 diminished both IFN-g production and cytolytic activity in 296 response to influenza-infected cells, but diminished only cytolytic activity, and not IFN-g production, in 297 response to K562 cells. This is likely due to a requirement for both cytokines (primarily type I interferons) 298 and cellular contact to obtain maximal IFN-g production by NK cells (Kronstad et al., in revision and 299 bioRxiv). While infected monocytes produce abundant IFN-a, K562 cells do not. 300

301
This study leaves open the question of the mechanism by which CD38 is increased during pregnancy. CD38 302 protein expression increases with age (39). Our pregnant and non-pregnant women were well-matched for 303 age, so this cannot account for the differences between cohorts. CD38 gene and protein expression is 304 regulated by transcription factors induced by diverse potential signals such as cytokines and hormones (40, 305 41). Thus, the altered hormonal and inflammatory environments during pregnancy could contribute to 306 enhanced CD38 expression. In particular, NK cells express the estrogen receptor beta (42). Estrogen binding 307 to its receptor in peripheral NK cells could increase CD38 expression during pregnancy, especially during 308 second and third trimester where a substantial increase in estrogen concentration in the blood is observed 309 (43). While this has not been explored in humans or in pregnancy, CD38 protein expression on 310 cardiomyocytes was increased following injection of systemic estrogen, but not progesterone, in a rat model 311 therapies. For immunotherapy approaches in which cytotoxic activity is desired, it will be important to 318 assure that these cells express CD38. Further, it will be important to understand the factors that control 319 CD38 expression in vivo to assure its retention on effector cells. For instance, we find that both IL-2 and 320 IL-15, two important cytokines for NK cell homeostasis, promote increased CD38 expression in vitro (data 321 not shown). It is equally important that we assess the role of CD38 on CD8 + cytotoxic T cells. Finally, our 322 data raise interesting implications for the CD38-targeting agents that are currently in clinical use, primarily 323 for multiple myeloma and chronic lymphocytic leukemia (46, 47). Clinical treatment with daratumumab, a 324 specific human CD38 binding antibody, improves patient outcomes (46), but also may lead to NK cell 325 fratricide as most NK cells express CD38 (48, 49). It is currently unclear how daratumumab affects CD38-326 mediated immune synapse formation, and how this influences NK cell killing. This will be an important 327 area of future investigation. 328

330
There are several limitations of our study, including the fact that our mass cytometry panels differed 331 between the two cohorts and remain limited to ~40 markers. Thus, we may have excluded other molecules 332 involved in NK cell immune responses during pregnancy, including critical NK cell surface molecules such 333 as DNAM-1, TIGIT and Siglec-7. We also did not follow-up on other differences that were seen in only 334 one cohort. Further, here we studied peripheral blood NK cells and were not able to sample lung resident 335 NK cells or uterine NK cells. Finally, we had limited data reflecting the prior history of the infected and 336 control women in terms of their prior vaccination status, prior influenza infection status, cigarette and drug 337 use, and others. We cannot exclude that unmeasured factors could influence the quality of the NK cell 338 response to influenza. The response ! specifies the experimental group for the th cell, e.g. encoding whether the cell was 394 stimulated ( ! = 0) or unstimulated ( ! = 1), and the explanatory variables are inverse hyperbolic sine 395 transformed protein counts ! . Each vector ! is of length equal to the number of measured proteins. Each 396 experiment will produce pairs of ( " , " ), … , ( # , # ) coming from different donors. Our statistical model 397 Mixed Model (GLMM). Using the GLMM terminology, the unknown parameters that we need to estimate 406 are the fixed effects , the random effects donor [!] , and the covariance matrix of the random effect. 407 GLMMs can account for donor specific heterogeneity by separating donor specific variation from overall 408 variation. As it is common in single cell datasets, we observe 10,000 to 100,000 cells per donor and measure 409 30 to 40 proteins. To handle such large amounts of data, we use the R package mbest that implements a fast 410 moment-based estimation to fit GLMM models (52). One key assumption in GLMMs is the modeling of 411 random effects by a multivariate normal distribution. 412 413

Generalized Linear Model with Bootstrap Resampling 414
Our GLMM approach can be directly applied to paired experiments. In immunology, we often have a paired 415 experimental design when comparing stimulated with unstimulated cells taken from the same blood sample. 416 We also have paired samples, if we compare the functional response of different cell types within the sample 417 blood sample. In contrast, in our pregnancy study, we have blood samples from pregnant and not pregnant 418 donors. Furthermore, we do not have any additional covariates that we could use to match samples. In such 419 cases, the GLMM approach might provide conservative results, therefore, we propose to use a 420 nonparametric bootstrap resampling (53) approach using standard Generalized Linear Models (GLMs) (54), 421 . 425

423
In this approach, we handle the donor heterogeneity without explicitly defining a parametric mixed effects 426 distribution. One bootstrap draw is taken by sampling donors with replacement. For each bootstrap sample, 427 we fit a GLM using the base R implementation. We repeat this procedure times resulting in coefficient  Following isolation or purification, NK cells were incubated for 5 min in Fc Block, washed, and incubated 467 for 15 min with CD38, CD31 or LFA-1 blocking antibodies or isotype control antibodies. NK cells were 468 then exposed to pH1N1-infected monocytes at a Effector:Target (E:T) ratio 1:1. Immediately following co-469 incubation, 2 µM monensin, 3 µg/mL brefeldin A, and anti-CD107a-allophycocyanin-H7 were added to the 470 co-culture for 4 hours, followed by cell staining for flow cytometry analysis.  Table S1. Demographics for cohort 1. 528 Table S2. Demographics for cohort 2. 529 Table S3. Antibody panel for mass cytometry in cohort 1. 530 Table S4. Antibody panel for mass cytometry in cohort 2. 531 532 Acknowledgements: We thank our study volunteers for their participation in these studies, Sally Mackey 533 for regulatory and data management, Sue Swope for consenting and conducting study visits and the staff of women (N=10) in cohort 1 were isolated from blood samples. Monocytes and NK cells were sorted and 679 monocytes were infected with the H1N1 influenza virus strain. NK cells were either exposed to H1N1-680 infected monocytes or to K562 tumor cells for 7h or 4h, respectively. The NK cell immune response was 681 PMBCs from controls and pregnant women in cohort 2 were isolated and labeled using a 31-parameter 720 antibody panel (see Table S3). CD38and CD38 + NK cells were separated by manual gating and analyzed 721 using GLM with bootstrap resampling (A) and conventional gating strategy for each marker (B, for 722 activating receptor expression, and C, for inhibitory receptor expression).