SYK negatively regulates ITAM-mediated human NK cell signaling and CD19-CAR NK cell efficacy

Abstract

Natural killer (NK) cells express activating receptors that signal through ITAM (immunoreceptor tyrosine-based activation motif)-bearing adapter proteins. The phosphorylation of each ITAM creates binding sites for SYK and ZAP70 protein tyrosine kinases to propagate downstream signaling including the induction of Ca²⁺ influx. While all immature and mature human NK cells coexpress SYK and ZAP70, clonally driven memory or adaptive NK cells can methylate SYK genes, and signaling is mediated exclusively using ZAP70. Here, we examined the role of SYK and ZAP70 in a clonal human NK cell line KHYG1 by CRISPR-based deletion using a combination of experiments and mechanistic computational modeling. Elimination of SYK resulted in more robust Ca²⁺ influx after crosslinking of the CD16 and NKp30 receptors and enhanced phosphorylation of downstream proteins, whereas ZAP70 deletion diminished these responses. By contrast, ZAP70 depletion increased proliferation of the NK cells. As immature T cells express both SYK and ZAP70 and mature T cells often express only ZAP70, we transduced the human Jurkat cell line with SYK and found that expression of SYK increased proliferation but diminished T cell receptor–induced Ca²⁺ flux and activation. We performed transcriptional analysis of the matched sets of variant Jurkat and KHYG1 cells and observed profound alterations caused by SYK expression. As depletion of SYK in NK cells increased their activation, primary human NK cells were transduced with a CD19-targeting chimeric antigen receptor and were CRISPR edited to ablate SYK or ZAP70. Deletion of SYK resulted in more robust cytotoxic activity and cytokine production, providing a new therapeutic strategy of NK cell engineering for cancer immunotherapy.

The Editors have selected this article as a highlight of the issue.

Introduction

Natural killer (NK) cells recognize and directly target cancer cells without the need for prior exposure. Unlike T cells, which require the presentation of specific antigens by major histocompatibility complex molecules, NK cells detect and eliminate cells that display stress-induced ligands, downregulation of major histocompatibility complex class I, or other markers associated with cellular transformation. NK cells use of germline-encoded receptor recognition of conserved ligands is advantageous in the context of cancer in which the heterogeneity of tumors makes it challenging to identify specific antigens that are universally expressed across different types of cancers. Chimeric antigen receptors (CARs), whether introduced into NK or T cells, share the same fundamental concept of being engineered to express synthetic receptors that recognize specific antigens on target cells and thus redirect killing. Development of CAR-NK cells is motivated by the unique characteristics of NK cells, such as their innate recognition abilities, broad specificity, and potential for reduced off-tumor toxicity.¹

NK cells provide rapid immune responses against viral infections and cancer. In addition to germline-encoded receptor recognition by NK cells, antibody-dependent cellular cytotoxicity (ADCC) is a mechanism that uses activating Fc receptors (FcγRIIIa, CD16a, FCGR3A) to provide exquisite antigen specificity. These receptors bind the Fc region of IgG and when bound to the surface of target cells can activate and release cytotoxic granules containing granzymes and perforin and secrete cytokines and chemokines.² Human NK cells that express CD16 are identified as the mature (CD56^dimCD16⁺) subset, whereas the lack of CD16 is a characteristic of the immature (CD56^brightCD16⁻) NK cell subset. Memory or adaptive NK cells have been identified in individuals exposed to human cytomegalovirus (CMV), and they possess the ability to undergo clonal expansion, generate long-lasting memory responses, and display enhanced effector functions including ADCC. In humans, these adaptive NK cells are identified by the surface expression of NKG2C and often the loss of expression of FcεR1γ, SYK, and/or EAT2.^3–6 Understanding the adaptive features of NK cells, particularly in the context of ADCC,⁷ has broad implications for cancer immunotherapy.

Immunoreceptor tyrosine-based activation motif (ITAM) signaling in NK cells provides activation through the CD16, NKp46, and NKp30 receptors and their associated adaptors FcεR1γ and CD3ζ, which form homo- or heterodimeric structures (e.g., FcεR1γ-FcεR1γ, FcεR1γ-CD3ζ, or CD3ζ-CD3ζ).² The NKG2C-CD94 complex is expressed on adaptive NK cells and noncovalently associates with DAP12 homodimers, which each carry 1 ITAM.^8,9 In a similar manner, T cells use the T cell receptor (TCR) complex with the associated CD3 adapters (CD3γ, CD3δ, CD3ε, and CD3ζ).¹⁰ Upon ligand binding, these receptors activate Src-family kinases (SFKs) such as Lck that promote the phosphorylation of ITAM residues. Phosphorylated ITAMs serve as docking sites for SYK and ZAP70 kinases, which possess tandem Src homology 2 (SH2) domains that specifically bind to these phosphorylated motifs. Once bound to the phosphorylated ITAMs, SYK and ZAP70 trigger pathways involving phospholipase C-γ (PLC-γ) leading to calcium release, PI3K (phosphoinositide 3-kinase) leading to AKT activation, and Vav1 leading to ERK activation.¹¹ Together, SYK and ZAP70 play crucial roles in transducing signals from ITAM-containing receptors that orchestrate NK and T cell immune responses.

SYK and ZAP70 belong to the family of nonreceptor tyrosine kinases and transmit signals downstream of immunoreceptors.^11,12 Both kinases share similar structures such as tandem SH2 domains at their N-terminus and a C-terminal kinase domain. SFKs phosphorylate tyrosine residues within their activation loop and cause a conformational change; however, only SYK can undergo autophosphorylation without the need of SFKs.¹¹ Furthermore, SYK and ZAP70 are differentially expressed in immune cell subsets. ZAP70 is predominately expressed in T cells, whereas SYK is expressed in B cells, mast cells, macrophages, dendritic cells, monocytes, and some γδ T cells.¹¹ Uniquely, most NK cells coexpress both SYK and ZAP70 but can downregulate SYK expression by methylation in adaptive NK cells.^3,4 Recent studies by Dahlvang et al.¹³ have reported that CRISPR-mediated ablation of SYK in primary human NK cells increases their cytolytic activity and cytokine production, whereas ablation of ZAP70 impairs these effector functions.

FcεR1γ and/or CD3ζ transduce signals via SYK and/or ZAP70, which makes studying these pathways and signaling molecules challenging. In this study, we have investigated the unique aspects of ITAM-induced activation by SYK and ZAP70 in human NK and T cells using a combination of experiments and mechanistic computational modeling.

Materials and methods

Human samples

Primary human NK cells were obtained from healthy donors with informed consent in accordance with approval by the University of California Committee on Human Research Institutional Review Board 10-00265. Peripheral blood mononuclear cells were obtained by Trima Residuals from apheresis collection of blood enriched for leukocytes, predominately mononuclear cells (Vitalant).

Cell lines

The human KHYG1 parental NK cell line¹⁴ was generously provided by Dr. Nicolai Wagtmann (Dragonfly Therapeutics).The human Jurkat leukemic cell line was generously provided by Dr. Arthur Weiss (University of California, San Francisco). HEK293T cells were obtained from the American Type Culture Collection (ATCC). The SUPB15 B lymphoblast cell line was obtained from the ATCC, transduced to express mKate2 and luciferase, and cultured in Iscove’s Modified Dulbecco’s Medium with 100 U/mL penicillin, 100 µg/mL streptomycin, 50 µM 2-mercaptoethanol, and 20% fetal calf serum (FCS). KHYG1, Jurkat, and HEK293T cells were cultured in complete RPMI 1640 (RPMI) or Dulbecco’s Modified Eagle Medium media supplemented with 2 mM glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, 50 µg/mL gentamicin, 110 µg/mL sodium pyruvate, 50 µM 2-mercaptoethanol, 10 mM HEPES, and 10% FCS. KHYG1 cells were supplemented with 200 U/mL recombinant human IL-2 (Teceleukin; provided by the National Cancer Institute Biological Resources Branch). Primary human NK cells were cultured in NK MACS medium (Miltenyi Biotec) supplemented with 5% human platelet lysate (EliteGro-Adv; Elite Cell), 0.5% penicillin-streptomycin (Gibco), and 1,000 U/mL of recombinant human IL-2 (PeproTech).

Antibodies and flow cytometry

Fluorochrome-conjugated antibodies were used to stain surface and intracellular proteins. Cells were first stained at 4 °C in Cell Staining Buffer (BioLegend) for 30 min, washed, and analyzed using a LSRII (BD Biosciences) or spectral Cytek Aurora (Cytek Biosciences). Intracellular staining required additional steps after surface staining using the BD Cytofix/Cytoperm Fixation/Permeabilization Solution Kit (BD Biosciences) following the manufacturer’s protocol. FcR were blocked using human TruStain FcX (BioLegend). Zombie Red (BioLegend) was used for Live/Dead exclusion dye. Data were analyzed using FlowJo software (TreeStar 10.8.1 and 10.10.0) by gating on lymphocytes identified based on forward and side light scatter properties and by excluding doublets, dead cells, and lineage-negative (CD14−CD19−) cells, as shown in Fig. 1. Anti-NKp30 (clone P30-15), anti-CD56 (clone NCAM16.2), anti-NKp46 (clone 9E2), anti-CD3 (clone UCHT1 and OKT3), anti-CD16 (clone 3G8), anti-CD57 (clone QA17A04), anti-Fc$\mathrm{\epsilon }$R1antibody $\mathrm{\gamma }$ subunit (product FCABS400F), anti-CD45 (clone 2D1), anti-SYK (clone 4D10.2), anti-PLZF (clone R17-809), anti-NKG2A (clone REA1100), anti-NKG2C (clone 134591), anti-ZAP70 (clone A151148 or 1E7.2), anti-CD19 (clone H1B19), anti-CD14 (clone 63D3), anti-CD69 (FN50), anti-CD247 (clone 6B10.2), anti-CD5 (clone UCHT2), anti-NGFR (clone C40-1457), anti-G4S Linker for CAR detection (clone E7O2V), anti-CD107a (clone H4A3), anti-TNF (Mab11), and anti-IFN$\mathrm{\gamma }$ (clone W19227A) were purchased from BioLegend, BD Biosciences, Cell Signaling Technology, Milli-Mark, or Thermo Fisher.

Figure 1.

SYK and ZAP70 expression profiling in adaptive NK cell subsets. (A–C) Healthy human donor peripheral blood mononuclear cells were analyzed by flow cytometry for surface and intracellular markers. (A) Gating strategy of immature (CD56^bright, CD16⁻), mature (CD56^dim, CD16⁺, FcεR1γ⁺, PLZF⁺), and adaptive (CD56^dim, CD16⁺, FcεR1γ⁻, PLZF-, NKG2C⁺) human NK cells. (B) Heatmap of mean fluorescence intensity (MFI) normalized by specific markers for each donor and gated on NK cell subsets. (C) Overlay histogram flow plots for each NK subsets showing SYK and ZAP70 expression. (D, E) Graphical representation of MFI and percentage data for SYK and ZAP70 for each human NK cell subset. Graphs show means with dots representing each donor; results from a paired-sample t test with significance are shown as *P < 0.05 and ***P < 0.001. Total of 10 biological samples were analyzed. AF647, Alexa Fluor 647; APC, allophycocyanin; BV, Brilliant Violet; PE, phycoerythrin; BUV, Brilliant Ultra Violet.

Genomic CRISPR-Cas9-based editing

CRISPR Cas9-RNP single guides were used to delete endogenous human SYK and ZAP70 genes from KHYG1 cells. CRISPR single guide sequence (5′-ACACCACTACACCATCGAGC-3′; IDT) was used to target exon 2 of human SYK and sequence (5′-TCATACACGCTCGTGTCCAT-3′; IDT) was used to target exon 2 of human ZAP70. In brief, KHYG1 cells were electroporated (4D-Nucleofector System; Lonza Bioscience) using the P3 Primary Cell Nucleofector Solution (Lonza Bioscience), 120 bp single-stranded DNA nontargeting electroporation stabilizer (IDT), and the corresponding CRISPR target guide RNA. KHYG1 cells were then single-cell sorted into 96-well plates and grown in RPMI 1640 media supplemented with 1,000 U/mL recombinant human IL-2 (PeproTech) for 5 to 6 wk. Intracellular flow cytometry was then used to screen for single-cell clones of either SYK- or ZAP70-depleted KHYG1 cells. Four SYK- and 4 ZAP70-ablated single cell clones were combined and used. Depletion of endogenous SYK or ZAP70 was confirmed using intracellular flow cytometry and Western blot analysis.

Retroviral transfection and transduction

Human SYK or CD16 complementary DNA sequences were cloned into pMXs-Puro retroviral vectors. Retrovirus were generated using HEK293T cells plated 1 d prior in 6-well plates and transfected using the Lipofectamine 2000 reagent according to the manufacturer’s protocol (Thermo Fisher Scientific), pMXs-Puro vector, and packaging plasmids. HEK293T viral supernatant was collected and used to transduced KHYG1 or Jurkat cells. Surface expression of CD16 was used to sort CD16-expressing KHYG1 cells. Jurkat cells were transduced with pMXs-Puro-SYK and then selected with puromycin for 5 to 6 wk. Expression of SYK in Jurkat cells was determined using intracellular flow cytometry with purity of >95% and ZAP70 expression matched that of the parental cells.

Western blotting

Cell protein was prepared using RIPA Lysis Buffer (Thermo Fisher Scientific) following the manufacturer’s protocol. Protein samples were first reduced using 355 mM 2-mercaptoethanol (Bio-Rad) in Laemmli Sample Buffer (Bio-Rad). Proteins were separated on 12% Mini-PROTEAN TGX precast protein gels (Bio-Rad) and then blotted to PVDF membranes using a 20% methanol pre–wet type blotting system and Criterion Blotter with Plate Electrodes (Bio-Rad). Membranes were washed with Tris-buffered saline with Tween 20 (TBST) (Bio-Rad) and incubated overnight at 4 °C or for 2 h at room temperature with blocking buffer (5% skim milk in TBST). The blots were washed and probed with 0.5 to 1 µg/mL of primary antibody against SYK (clone 4D10.2; BioLegend) or ZAP70 (clone 1E7.2; BioLegend) for 1 to 2 h with gentle rocking. Protein loading control horseradish peroxidase was detected using anti-GAPDH (clone W17079A; BioLegend). The blots were then washed 3 times with TBST and incubated with a secondary antibody, horseradish peroxidase–conjugated goat anti-mouse IgG (BioLegend), for 1 h at room temperature. Membranes were incubated in chemiluminescent substrate (SuperSignal West Pico Plus; Thermo Fisher Scientific) for 5 min and visualized using a ChemiDoc Imaging System (Bio-Rad).

Phosphorylation kinetics assay

Cells were stimulated by antibody-induced receptor crosslinking, probed with multiple phosphorylated antibodies, and analyzed by flow cytometry. Cell numbers were first normalized between groups by using the Countess 3 FL Automated Cell Counter (Thermo Fisher Scientific). A total of 1 to 2 million cells were stained with 5 µg/mL of biotinylated anti-CD16 (clone 3G8; BioLegend), anti-CD3 (clone UCHT1; BioLegend), or anti-NKp30 (clone P30-15; BioLegend), and Zombie Red Fixable Viability dye (BioLegend) depending on the cell type or activation and incubated for 30 min at 4 °C in Hank’s Balanced Salt Solution (Thermo Fisher Scientific) Stimulation media (Hank’s Balanced Salt Solution with Ca²⁺ and Mg²⁺, no phenol red, 10 mM HEPES, and 1% FCS). The cells were washed and rested at 37 °C for 60 min. Receptor were then crosslinked by the addition of streptavidin (500 µg/mL; BioLegend) and quenched at the conclusion of the indicated time points. The stimulation was quenched with the addition of Fixation buffer (BD Biosciences) and incubated for an additional 15 min at 37 °C. The cells were permeabilized using the True-Phos Perm Buffer (BioLegend) according to the manufacturer’s protocol. Cells were stained with anti-LYN (Tyr397)/LCK (Tyr394)/HCK (Tyr411)/BLK (Tyr389) (clone E5L3D; Cell Signaling Technology), anti-LAT (pY226) (clone A20005; BioLegend), anti-ZAP70 (Tyr493) (clone A16043E; BioLegend), anti-ZAP70 (Tyr292) (clone A16038B; BioLegend), anti-MAPK (Thr180/Tyr182) (clone A16016A; BioLegend), anti-ERK1/2 (Thr202/Tyr204) (clone 6B8B69; BioLegend), anti-AKT (pS473) (clone M89-61, BD Bioscience), anti-SLP-76 (Ser376) (clone E3G9U; Cell Signaling Technology), and anti-PLC$\mathrm{\gamma }$1 (Tyr783) (clone A17025A; BioLegend) for 30 min at room temperature. The cells were then analyzed on the spectral Cytek Aurora analyzer (Cytek Biosciences).

Calcium influx assays

T and NK cells were labeled with 7 µg/mL of Indo-1 AM (Thermo Fisher Scientific) for 30 min, washed, and stained for an additional 30 min at 4 °C either with biotinylated anti-CD16 (clone 3G8; BioLegend) or anti-CD3 (clone UCHT1; BioLegend). Cells were then stimulated by crosslinking the receptors with mAb and streptavidin in a manner similar to phosphorylation kinetic assays (BioLegend). Cells were analyzed on the flow cytometer with ultraviolet fluorescence channel detectors by acquiring the first 30 s of baseline calcium levels and then measuring calcium influx levels upon the addition of streptavidin (500 µg/mL; BioLegend) while maintaining 37 °C temperature using a water pump warmer.

Primary CAR-NK and T cell engineering

The cell engineering platform for expansion and genome editing of primary human NK cells were adapted from Huang et al.¹⁵ Primary human NK cells were isolated from peripheral blood mononuclear cells obtained after density gradient centrifugation of Trima Residuals from apheresis collection (Vitalant). Negative selection of NK cells was performed by using a EasySep Human NK Cell Enrichment Kit (STEMCELL). Cells were cultured in NK MACS medium (Miltenyi) supplemented with recombinant human IL-2 1000 U/mL (PeproTech). Cells were activated over 7 d with anti-CD2 and anti-NKp46-coated beads (Miltenyi Biotec) and cultured at 1 × 10⁶ cells/mL. At day 7, beads were removed and NK cells were transduced by centrifugation on Retronectin (Takara)-coated plates with a concentrated (Retro-X concentrator; Takara) SFG γ-retroviral vector encoding a second-generation anti-CD19 CAR with CD28 costimulation and CD3ζ domains packaged using 293Vec-RD114 cells (Biovec Pharma). Genes of interest were subsequently deleted by CRISPR editing in transduced NK cells. After 2 d of rest in complete NK cell medium, 2 × 10⁶ NK cells were electroporated with 40 pmol of purified Cas9-NLS protein (MacroLab) complexed with 80 pmol of sgRNA (Synthego) using a 4D nucleofector (Lonza Bioscience) in 20 μL of P3 buffer and supplement (Lonza Bioscience) using the CM-137 pulse code. Immediately after electroporation, edited NK cells were rescued in prewarmed culture medium and plated for further cell culture. Primary T cells were cultured in X-VIVO 15 media (Lonza Bioscience) supplemented with 5% human serum, 5 ng/mL human IL-7, and 5 ng/mL human IL-15 (Miltenyi Biotec) and 0.5% penicillin-streptomycin (Gibco). Briefly, T cells were activated using anti-CD3+CD28 beads (Thermo Fisher Scientific) with a 1:1 bead-to-cell ratio. After 48 h of stimulation, the beads were removed, and the T cells were retrovirally transduced to express a CD19-targeting CAR as described previously. The following day, cells were transduced with the SYK-expressing vector. The next day, T cells were edited with CRISPR-Cas9 RNP targeting ZAP70 or AAVS1 as a negative control. Cells were allowed to rest for 5 d before conducting experiments.

Proliferation assay

CAR T cells were labelled with CellTrace Violet (Thermo Fisher Scientific) according to manufacturer’s recommendation. Briefly, cells were incubated with diluted CellTrace Violet in phosphate-buffered saline and stained for 30 mins at 37 °C. Cells were then thoroughly washed with media, plated at density of 50,000 cells per well of 96-well round bottom plate in 200 µL of media containing IL-7 and IL-15. Cells were incubated at 37 °C incubator and analyzed 3 d later using flow cytometry while simultaneously staining for NGFR, SYK, and ZAP70.

Flow cytometry–based 4-h cytotoxicity assay

The cytotoxicity of NK cells in a 4-h assay was measured by flow cytometry as adapted from Aguilar et al.¹⁶ In brief, CD19-expressing SUP-B15 (ATCC) cells were labeled with CellTrace Violet and cocultured with primary CAR-NK cells at varying effector-to-target ratios. Cells were cultured for 4-h and then washed and stained with Zombie Red Fixable Viability dye (BioLegend). CountBright Absolute Counting Beads (Thermo Fisher Scientific) were added to each well to obtain cell counts according to the manufacturer’s protocol. Percentage of specific target cell lysis was calculated with the following formula, which determines the number of live cells in experimental samples relative to control samples:

$$\mathrm{\%}\mathrm{ }\mathrm{specific}\mathrm{ }\mathrm{lysis}\mathrm{ }=\mathrm{ }100\mathrm{ }\mathrm{x}\mathrm{ }\left(\frac{\#\mathrm{ }of\mathrm{ }\mathit{live}\mathrm{ }\mathit{cells}\mathrm{ }in\mathrm{ }\mathit{experimental}-\#\mathrm{ }of\mathrm{ }\mathit{cells}\mathrm{ }in\mathrm{ }\mathit{spontaneous}\mathrm{ }\mathit{control}}{\#\mathrm{ }of\mathrm{ }\mathit{live}\mathrm{ }\mathit{cells}\mathrm{ }in\mathrm{ }\mathit{maximal}\mathrm{ }\mathit{control}-\#\mathrm{ }of\mathrm{ }l\mathrm{i}ve\mathrm{ }\mathit{cells}\mathrm{ }in\mathrm{ }\mathit{spontaneous}\mathrm{ }\mathit{control}}\right)$$

Luciferase-based 24-h cytotoxicity assay

The cytotoxicity of NK cells was determined by using a standard luciferase-based assay. In brief, SUP-B15 cells were transduced with lentivirus to express firefly luciferase-mKate2 and served as target cells. The effector and tumor target cells were cocultured in triplicates at the indicated effector-to-target ratios using white-walled 96-well flat clear-bottom plates with 5 × 10⁴ target cells in a total volume of 100 μL per well in NK cell medium without IL-2. The control for maximal signal was SUP-B15 cells alone, and the control for minimum signal was SUP-B15 cells with Tween 20 (0.2%). Cocultures were incubated for approximately 20 to 24 h. Then, 100 μL D-luciferin (GoldBio; 0.75 mg/mL) was added to each well and the luminescent signal was measured using a GloMAX Explorer microplate reader (Promega).

$$\mathrm{\%}\mathrm{ }\mathrm{specific}\mathrm{ }\mathrm{cytotoxicity}=100\mathrm{ }x\mathrm{ }1-\frac{\left(\mathit{sample}-\mathit{minimum}\right)}{\left(\mathit{maximum}-\mathit{minimum}\right)}.$$

Next-generation sequencing

Total RNA was extracted from cells using the Qiagen RNeasy Mini kit with DNase treatment per the manufacturer’s protocol. Illumina libraries were generated from total RNA using the Universal Plus mRNA-Seq with NuQuant kit and sequenced on the HiSeq to a depth of $\ge$66 million reads. Sequences were align to the Ensembl human genome build GRCh38 annotation version 95 using STAR version 2.7.5c.¹⁷ Differentially expressed genes were analyzed between groups using DESeq2 version 1.42.1¹⁸ on R software version 4.3.3 (R Foundation for Statistical Computing). RNA sequencing data were deposited in the National Center for Biotechnology Information Sequence Read Archive access number PRJNA1171962 (https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1171962).

Graph statistical analysis

Data were analyzed using Prism version 10.2.3 (GraphPad Software) by unpaired and paired t test. Graphs show mean ± SD. Additional statistical analysis details can be found in the figure legends.

In silico modeling membrane proximal CD16 signaling

We developed a rule-based multicompartment model describing CD16 receptor–initiated membrane proximal biochemical signaling reactions in which the participating molecules are well mixed.¹⁹⁻²¹ The model is composed of CD16 receptor, anti-CD16 antibody, CD3ζ adaptor, SFK Lck, SYK-family kinases ZAP70 and SYK, phosphatase SHP-1, and ubiquitin ligase Cbl.^22–26 The signaling protein species are placed within a 2 compartmental simulation box representing thin layers of the extracellular region (compartment 1) of depth 2 nm and the cytosolic region (compartment 2) of depth 1 μm partitioned by the 2-dimensional NK cell plasma membrane of area 5 μm × 5 μm. Plasma membrane bound CD16 binds with the Fc region of the antibody ligands in compartment 1, while the transmembrane domain of CD16 binds with various signaling molecules in compartment 2. CD16, Lck, and CD3ζ adaptors are always anchored to the membrane. ZAP70, SYK, and SHP-1 are cytosolic molecules that bind to the ITAM present on the tails of the CD3ζ proteins extended in the cytosol. Cbl, a cytosolic ubiquitin ligase, binds to the ITAM through ZAP70 or SYK phosphorylation.^27,28 The major signaling reactions in the simulation box are shown in Fig. 2G and discussed in our GitHub link. The details of the signaling reactions, the associated rates, and abundances (or copy numbers) of CD16, antibody ligands, and signaling proteins used in the model are shown in Table S1. Each ITAM on CD3ζ can exist in 10 distinct states including fully and partially unphosphorylated states, fully phosphorylated states, and fully phosphorylated states bound to ZAP70,²⁹ and fully or partially phosphorylated states bound to SYK and/or SHP-1, respectively.^30,31 Because homodimers of CD3ζ adaptor have 6 ITAMs, it can generate up to ∼6¹⁰ possible intermediate complexes in the model and increases computational runtime significantly when ordinary differential equations (ODEs) describing mass action rate kinetics are used. To save computational cost, we simulated the biochemical signaling reaction kinetics using stochastic NFSIM (Network Free Simulator) application.^21,32 The NFSIM application simulates the kinetics with intrinsic noise fluctuations using Gillespie’s algorithm.²⁰

Figure 2.

SYK and ZAP70 signaling kinetics in KHYG1 NK cells. (A–E) KHYG1 cells were transduced with a retroviral vector expressing human CD16 and then depleted (Δ) of either SYK or ZAP70. (A) Diagram of biological regulation of SYK and ZAP70. (B, C) SYK and ZAP70 ablation was confirmed by Western blot and flow cytometric analysis. (D) Calcium influx assay using SYK- and ZAP70-ablated KHYG1 cells. Cells were labeled with indo-1, stained with biotinylated anti-CD16 antibody, and then crosslinked with streptavidin. Cells were collected on a flow cytometer continuously for the first 30 s without crosslinking and then crosslinked for an additional 270 s. (E) Graphical representative of the calcium influx for the SYK- and ZAP70-ablated KHYG1 cells; the area under the curve (AUC) was calculated for each cell line, with replicates. (F) Cell proliferation was measured for 96 h in media with 200 U/mL IL2. Sample t tests with significance are shown as ****P < 0.0001. Data in panels D and E represent 3 to 4 biological replicates for 2 independent experiments. (G) Graphical representation of the major CD16 initiated signaling events involving CD16, Lck, ZAP70, SYK, and SHP-1 in the simulation box (left) and the ITAM phosphorylation by SYK and autophosphorylation and transphosphorylation of SYK (right). (H) Graphical representative of the mechanism of increased competition between ZAP70 and SYK for accessing phosphorylated ITAMs in WT NK cells for limited number of ITAMs. (I) Shows model fits (solid lines) for Ca²⁺ kinetics from 0 s to 240 s with the experimental counterparts for ΔSYK (red circles) and ΔZAP70 (blue circles) KHYG1 cells. The predicted Ca²⁺ kinetics (solid black line) for WT NK cells by the trained in silico model shows an excellent agreement (R² = 0.82) with the Ca²⁺ kinetics measured in WT KHYG1 cells (black circles). (J) The number of ITAM-bound ZAP70 (left) and ITAM-bound SYK (right) in the simulation box predicted by the trained in silico model for WT, ΔSYK, and ΔZAP70 cells. PE, phycoerythrin; UV, ultraviolet.

Minimal model for Ca²⁺ signaling

We developed a minimal model for Ca²⁺ kinetics initiated by phosphorylated forms of ZAP70 and SYK in the NK cell by modifying a coarse-grained model of Ca²⁺ kinetics in Xenopus oocyte cells.³³ In our model, the Ca²⁺ kinetics for pZAP70 or pSYK are given by 2 coupled nonlinear ODEs in terms of the concentration ([Ca²⁺]) of Ca²⁺ in the cytosol and the dimensionless variable h, which represents the proportion of activated (open channel) IP3 receptors on the endoplasmic reticulum (ER) that release Ca²⁺from ER. The ODEs contain parameters b, k₁, and k₂, which are related to the rates of Ca²⁺ release through IP3 receptors from the ER to the cytosol in the basal (parameter b) and activated state (parameters k₁ and k₂),^33,34 and γ, which is related to the rate of Ca²⁺ pumped out of the cytosol (Table S1). In the ODEs, the average concentrations of pZAP70 [pZAP70] and pSYK [pSYK] at any time t regulate the production of cytosolic Ca²⁺ and the proportion of activated IP3 receptors. Because there are differences in signaling reactions downstream of pZAP70 and pSYK that give rise to Ca²⁺ flux, we choose different values for some of the parameters for pZAP70- and pSYK-induced Ca²⁺ flux kinetics model. The ODEs for cytosolic ${\left[C{a}^{+2}\right]}_i$ induced by concentrations pZAP70 (${\left[\mathit{pZAP}70\right]\left(t\right)\equiv y}_{Z }\left(t\right)$) or pSYK (${\left[\mathit{pSYK}\right]\left(t\right)\equiv y}_{S }\left(t\right)$) are given below. Here, subscript i = Z (or S) denotes stimulation by pZAP70 (or pSYK).

$$\frac{d{\left[C{a}^{+2}\right]}_i}{dt}=C_{1i}{h}_i{y}_{i }\left(t\right) \frac{b{k}_1+{\left[C{a}^{+2}\right]}_i}{k_1+{\left[C{a}^{+2}\right]}_i}-{\gamma }_i{\left[C{a}^{+2}\right]}_i$$ (1)

$$\frac{d{h}_i}{dt}=C_{2i}{y}_{i\mathrm{ }}\left(t\right)(\frac{{k_2}^2}{{k_2}^2+{{\left[C{a}^{+2}\right]}_i}^2}-h_i)$$ (2)

We used different parameter sets {C_1Z, C_2Z, γ_Z} and {C_1S, C_2S, γ_S} for pZAP70- and pSYK-induced Ca²⁺ flux kinetics model (see estimated values in Table S1). When both ZAP70 and SYK are present in the cell, the total cytosolic Ca²⁺ concentration is given by, ${\left[C{a}^{+2}\right]}_{\mathit{total}}={\left[C{a}^{+2}\right]}_Z+{\left[C{a}^{+2}\right]}_S$ (see Fig. S1A). In experiment, Ca²⁺ flux (Fig. 2E) is measured in the unit of fluorescence ratio [Ca²⁺ = UV C (high Ca²⁺)/UV C (low Ca²⁺)], which we assumed to be proportional to the cytosolic Ca²⁺ concentration ${\left[C{a}^{+2}\right]}_{\mathit{total}}$, where UV C (low Ca²⁺) represents the basal value. For simplicity, we assumed the proportional factor f as 1 with a unit of [μM]⁻¹.

Preprocessing data for parameter estimation

We excluded the Ca²⁺ flux data for ΔSYK, ΔZAP70, and wild-type (WT) KHYG1 cells during the resting period (30 s) before stimulation. Therefore, in our in silico analysis, we modeled the Ca²⁺flux kinetics between 30 s and 270 s, as shown in Fig. 2E. The starting time, t = 0, of the model simulation (Fig. 2I) thus corresponds to t = 30 s in Fig. 2E.

Pipeline for parameter estimation

We developed a rule-based CD16 signaling model to generate pZAP70 or pSYK signal. We first created the reaction rules of CD16 signaling model using the Virtual Cell (version 7.5) software and exported the signaling model in .bngl (BioNetGen) format to run in Python using PyBioNetGen (PyBNG) library.^20,35 In the .bngl file (see our GitHub link), we obtained the copy number of signaling molecules in the simulation box by multiplying the species concentrations with area or volume with given in Table S1. Similarly, we used microscopic reaction rates (in the units of seconds and molecules) for the second-order biochemical reactions occurring either in the membrane or in volume (cytosolic or extracellular). We derived these microscopic rates by scaling the second order reaction rates (in units of μm³ s⁻¹ molecule⁻¹ or μm² s⁻¹ molecule⁻¹) in Table S1 with the volume (or area) of reactions that occurs in different compartments in the model (Table S1), namely, the extracellular volume (V_e), the intracellular cytosolic volume (V_c), and the membrane area (A). The simulation box represents a small transmembrane region in the NK cell, and we assumed that the pZAP70 and pSYK abundances in an NK cell are generated from the averages of the copy numbers from many such simulation boxes. Thus, the average of the pZAP70 and pSYK abundances were generated from independent runs of the model. We calculated the average of the copy numbers of pZAP70 and pSYK in the simulation box (or V_c) over a few (∼3) numbers of stochastic trajectories (Fig. 2J; Fig. S1B, C). For simplicity, we assumed that all the NK cells in the experiments have similar mean copy numbers of signaling proteins and ignored the cell-cell variation in the copy numbers of the proteins. Then, we input the average pZAP70 or pSYK (in concentration μM unit) in equations (1) and (2) to obtain cytosolic Ca²⁺ flux averaged across the cell population. We estimated parameters (see Table S1) by fitting the model generated Ca²⁺ flux with experimental Ca²⁺ flux for ΔSYK and ΔZAP70 KHYG1 cells simultaneously (Fig. 2I). For that, we minimized a combined cost function (defined as cost function = RSS_ΔZAP70 + RSS_ΔSYK), which represents the sum of individual residual sum of squares (RSS)³⁶ between model fit and experimental Ca²⁺ flux for ΔSYK and ΔZAP70 KHYG1 cells given by RSS = ${\sum }_t{\left({{Ca}^{++}}_{\mathit{model}}\left(t\right)-{{Ca}^{++}}_{\mathit{expt}.}\left(t\right)\right)}^2$, summed at times = 0 s to 240 s in the interval of 1 s (Fig. 2I). We employed the particle swarm optimization technique for parameter estimation using the pyswarm library using biologically relevant parameter ranges (upper and lower bounds). For particle swarm optimization execution, we choose the swarm size of particles and the number of iterations to be 40 and 20, respectively.

Predicting Ca²⁺ flux for WT KHYG1 NK cell

We considered the CD16 signaling model with ZAP70 and SYK (details of reaction rules are available in our GitHub link and Table S1) to describe CD16 signaling in WT NK cells. Then we calculated the average pZAP70 and pSYK abundances in WT NK cells using the set of parameters (Table S1) estimated from the in silico model. Finally, we predicted the cytosolic Ca²⁺flux assuming the Ca²⁺ flux is the superposition of the Ca²⁺ fluxes induced by pZAP70 and pSYK, i.e., ${\left[C{a}^{2+}\right]}_{\mathit{total}}={\left[C{a}^{2+}\right]}_{Z\mathrm{AP}70}+{\left[C{a}^{2+}\right]}_{\mathit{SYK}}$ (see Fig. S1A), where ${\left[C{a}^{2+}\right]}_{Z\mathrm{AP}70}$ and ${\left[C{a}^{2+}\right]}_{\mathit{SYK}}$ are the Ca²⁺flux stimulated by pZAP70 and pSYK, respectively. We assumed the same initial (t = 0) flux of [Ca²⁺] for ${\left[C2\right]}_{Z\mathrm{AP}70}$ and ${\left[C{a}^{2+}\right]}_{\mathit{SYK}},$ which is half of the experimental Ca²⁺ flux value at t = 0 s (Fig. 2I).

Parameter estimation

We first developed rule based CD16 signaling model in the Virtual Cell software.^19–21 Then exported the signaling model in BioNetGen format (.bngl) and computed the average of the stochastic kinetics of the concentrations of pZAP70 and pSYK in the CD16 signaling model using PyBioNetGen library in Python.³⁵ Then, we used the previous time courses of pZAP70 or pSYK as input in the ODEs in equations (1) and (2) and solved the ODEs numerically using the odeint function in SciPy Integrate Python library. We set up a cost function describing the RSS³⁶ in the ratio of the Ca²⁺ flux estimated in the model and the values measured in experiments for ΔSYK and ΔZAP70 KHYG1 NK cells (Fig. 2E). We estimated parameter values (see Table S1) by minimizing the cost function in Python using Particle Swarm Optimizer library pyswarm (https://pythonhosted.org/pyswarm/).

Calculating R²

To show goodness of our in silico model prediction, we calculated

$$\begin{array}{c}\mathrm{t}\mathrm{he}\mathrm{ }\mathrm{coefficient}\mathrm{ }\mathrm{of}\mathrm{ }\mathrm{determination}\mathrm{ }\left({\mathit{R}}^2\right)\mathrm{ }=\mathrm{ }(1-\frac{\mathit{sum}\mathrm{ }\mathit{squared}\mathrm{ }\mathit{regression}\mathrm{ }\left(\mathit{SSR}\right)}{\mathit{total}\mathrm{ }\mathit{sum}\mathrm{ }of\mathrm{ }\mathit{squares}\mathrm{ }\left(\mathit{SST}\right)})\\ =\left(1-\frac{{\sum }_t{\left({{Ca}^{++}}_{\mathit{expt}.}\left(t\right)-{{Ca}^{++}}_{\mathit{model}\_\mathit{prediction}}\left(t\right)\right)}^2}{{\sum }_t{\left({{Ca}^{++}}_{\mathit{expt}.}\left(t\right)-\overline{{{Ca}^{\pm +}}_{\mathit{expt}.}\left(t\right)}\right)}^2}\right)\end{array}$$ (37)

between Ca²⁺ flux observed in experiments (${{Ca}^{2+}}_{\mathit{expt}.}\left(t\right)$) and predicted through the in silico model (${{ Ca}^{2+}}_{\mathit{mode}{l}_{\mathit{prediction}}}\left(t\right)$) in WT KHYG1 NK cells (Fig. 2I) in time. The t subscript indicates the time index representing Ca²⁺ flux values measured at discrete time points between 0 to 240 s at the interval of 1 s. $\overline{{{Ca}^{2+}}_{\mathit{expt}.}\left(t\right)}$ represents the average of Ca²⁺ flux in experiments (${{Ca}^{2+}}_{\mathit{expt}.}\left(t\right)$) over time.

Details of the signaling model and availability of codes

Codes are written in Python programming language. The reaction rules for the rule-based CD16 signaling model can be visualized in the Virtual Cell software. The details of the signaling model and codes are available at the GitHub link: https://github.com/indraniny/ZAP_SYK_roles.

Results

Assessment of SYK and ZAP70 in NK cell subsets

Prior studies have demonstrated coexpression of SYK and ZAP70 in human NK cells. Here, we determined the relative expression of these kinases in immature (CD56^bright, CD16⁻), mature (CD56^dim, CD16⁺), and adaptive NK cells expressing NKG2C and lacking FcεR1γ and PLZF. We analyzed 10 total donors: 5 donors (1–5) had abundant adaptive NK cells and 5 donors (6–10) had very few adaptive NK cells (Fig. 1A). Using 15-parameter quantitative spectral cytometry, we confirmed the distinct phenotypic signature of the 3 major NK cell subsets found in the peripheral blood (Fig. 1B). In all donors, there was progressively less SYK in the transition from immature to mature, and the lowest expression of SYK was observed in the adaptive NK cells. By contrast, ZAP70 expression increases in the immature to mature state and maintained at high levels in adaptive NK cells (Fig. 1C, D). No significance was observed for ZAP70 percentages between the NK subset, and adaptive NK cells were absent in SYK (Fig. 1E). Therefore, we further investigated if these differences in signaling molecules affect the function of adaptive NK cells.

Ablation of SYK and ZAP70 in human NK cells

To investigate the role of SYK and ZAP70 in NK cell activation we selected a homogeneous clonal human NK cell line, KHYG1,¹⁴ which is typical of mature NK cells, that coexpresses SYK and ZAP70. Similar to some adaptive NK cells, KHYG1 expresses CD3ζ adaptor, but not FcεRIγ. As KHYG1 cells do not express CD16, they were transduced with CD16 to evaluate ITAM-based signal transduction through the activating CD16-CD3ζ receptor. Either SYK or ZAP70 was ablated by CRISPR editing technology (Fig. 2A). Complete ablation of SYK or ZAP70 was confirmed by Western blot analysis and flow cytometry (Fig. 2B, C). We measured Ca²⁺ influx in WT, SYK-ablated ($\Delta$SYK), and ZAP70-ablated ($\Delta$ZAP70) indol-1–labeled KHYG1 cells stimulated through CD16 by first coating cells with biotin-conjugated anti-CD16 and then crosslinking the receptor with streptavidin while measuring kinetics on a flow cytometer. Ablation of SYK dramatically increased the magnitude of the Ca²⁺ influx in response to CD16 stimulation, whereas ablation of ZAP70 diminished the response compared with WT NK cells (Fig. 2D, E). In contrast to the enhanced CD16-induced Ca⁺⁺ response of $\Delta$SYK KHYG1 cells, proliferation of $\Delta$ZAP70 KHYG1 was remarkably increased over the course of 4 d in medium with IL-2 (which is required for growth and viability of NK cells) compared with WT or $\Delta$SYK KHYG1 cells (Fig. 2F).

In silico modeling of signaling in WT, SYK-deficient, and ZAP70-deficient NK cells

We set up an in silico model to investigate the mechanistic roles of SYK and ZAP70 in regulating Ca⁺⁺ kinetics in response to CD16 stimulation in KHYG1 cells. In the in silico model, CD16 receptors bind to cognate ligands, and then homodimers of CD3ζ adaptors bind to the transmembrane domain of the ligand-bound CD16 receptors (Fig. 2G). ITAMs associated with CD3ζ adaptors are phosphorylated by membrane bound SFKs. SYK can bind to partially and fully phosphorylated ITAMs (Fig. 2G, right).^30,31 In contrast, ZAP70 can only bind to fully phosphorylated ITAMs.²⁹ Moreover, unlike ZAP70, SYK when bound to ITAMs can phosphorylate tyrosine residues in the ITAMs of the same CD3ζ or ITAMs of other CD3ζ proteins. The phosphorylation of ITAMs by SYK in turn can lead to increased recruitment of SYK to the ITAMs, which constitutes a positive feedback.³¹ ITAM-bound SYK can also autophosphorylate or transphosphorylate tyrosine residues in kinase domains on the same SYK or other SYK molecules, respectively, which increases the catalytic activity of SYK.³¹ The previous signaling reactions involving SYK thus can give rise to positive feedback in producing catalytically active SYK molecules. Phosphorylated ZAP70 and SYK molecules are dephosphorylated by phosphatases including SHP-1.³⁷ In addition, the ubiquitin ligase Cbl can bind to SYK and ZAP70 complexed with CD3ζ and CD16 and degrade the entire complex.^27,28 We modeled the stimulation of intracellular Ca⁺⁺ flux by phosphorylated ZAP70 and SYK minimally using a coarse-grained model described by ODEs in which the abundances (copy numbers per cell) of the phosphorylated forms of ZAP70 and SYK give rise to Ca⁺⁺ flux (see Materials and Methods for details). First, we trained our in silico model with time-dependent Ca⁺⁺ flux in the ΔZAP70 and ΔSYK cells (Fig. 2I) and estimated several model parameters. Then, we validated our in silico model by predicting Ca²⁺ flux kinetics in the WT KHYG1 cells, which shows excellent agreement (R²  = 0.82) with the data (Fig. 2I). The model determined the following mechanism for competition between ZAP70 and SYK (Fig. 2H). The intermediate magnitude of Ca²⁺ flux in the WT KHYG1 cells compared with the lower Ca²⁺ flux in the ΔZAP70 KHYG1 cells and higher Ca²⁺ flux in ΔSYK KHYG1 cells point to a competition between ZAP70 and SYK for access of the phosphorylated ITAMs in the WT KHYG1 cells. As ZAP70 and SYK exhibit similar affinity for binding to fully phosphorylated ITAMs,³⁸ the abundances of ITAM-bound ZAP70 and SYK generated during early time signaling in WT KHYG1 cells were reduced compared with their counterparts in ΔSYK and ΔZAP70 KHYG1 cells (Fig. 2J). Consequently, this results in an intermediate flux of Ca²⁺ in the WT KHYG1 in between the Ca²⁺ flux observed in ΔSYK and ΔZAP70 KHYG1 cells (Fig. 2I). The Ca²⁺ flux in the WT cells is determined largely by pZAP70 (Fig. S1A). Furthermore, the model predicts that when the abundances of CD16 and CD3ζ adaptors are in excess compared with that of ZAP70 and SYK in the WT NK cells, the competition between ZAP70 and SYK for binding the phosphorylated ITAMs decreases (Fig. S1E, F, J–L)). The positive feedback in SYK activation along with this effect produces a greater number of pSYK and pZAP70 in the WT cells compared with that of the ΔZAP70 and ΔSYK NK cells (Fig. S1H, I). Consequently, this leads to a substantially larger Ca²⁺ flux in the WT NK cells, prevailing over that of the ΔSYK and ΔZAP70 NK cells (Fig. S1D).

SYK depletion increases activation-induced phosphorylation kinetics in KHYG1

The kinetics in the phosphorylation of key signaling proteins downstream of ITAM-based signaling were analyzed by multiparameter spectral cytometry (Fig. 3A, B). We observed a significant increase in the phosphorylation kinetics for SYK-ablated KHYG1 cells with a peak phosphorylation of ERK1/2 (Thr202/Tyr204) at 2 min when crosslinked with either anti-NKp30 or anti-CD16 antibodies (Fig. 3C). Remarkably, ablation of SYK in KHYG1 cells stimulated through CD16 demonstrated a significant increase in many other phosphorylated proteins, including LYN (Tyr397), LCK (Tyr394), HCK (Tyr411), BLK (Tyr389), LAT (Tyr171), ZAP70 (Tyr493), MAPK (T180/Y182), ERK1/2 (Thr202/Tyr204), AKT (Ser473), and PLCy1 (Tyr783) when compared with WT KHYG1 cells (Fig. 3D, E). Phosphorylation of the tyrosine residue 292 on ZAP70 (Tyr292), which functions as a negative regulator, and SLP76 (Ser376) between ZAP70-ablated and SYK-ablated KHYG1 cells were unchanged (Fig. 3D, E).³⁹ Additionally, ZAP70-deficient KHYG1 cells showed multiple diminished phosphorylation events compared with WT KHYG1 cells except for the combined LYN (Tyr397), LCK (Tyr394), HCK (Tyr411), BLK (Tyr389), PLCy1 (Tyr783), and MAPK (T180/Y182) (Fig. 3D, E). SYK-ablated cells increased many phosphorylation events of ITAM-related proteins, whereas ZAP70-ablation decreased phosphorylation in NK cells.

Figure 3.

SYK depletion increases activation induced phosphorylation kinetics in KHYG1 cells. (A–D) Multiphosphorylation signaling markers were analyzed by flow cytometry for each NK cell line. KHYG1 cells were transduced with CD16 and then ablated (Δ) of SYK or ZAP70. (A) Diagram of KHYG1 cells stained with biotinylated anti-CD16 antibody and then crosslinked with streptavidin for 0 (unstimulated), 1, 2, 5, 10, 15, and 30 min and analyzed by spectral cytometry. (B) Representative data of phosphorylated pERK1/2 at Thr202 and Tyr204 over 30 min for each KHYG1 cell line. (C) Summarized data of pERK1/2 after crosslink activation with anti-NKp30 (left) and anti-CD16 (right). (D) Heatmap of geometric mean fluorescence intensity (MFI) values overtime and normalized to rows for each phosphorylated marker. (E) Line plot graph of data from cells crosslinking using anti-CD16. Data used for panels B, C, and D consist of 3 biological replicates. *P < 0.05. BV, Brilliant Violet.

Effect of SYK expression on T cell signaling and proliferation

Like NK cells, immature T cells in the thymus coexpress SYK and ZAP70, but SYK is downregulated as T cells mature,⁴⁰ which is similar to the downregulation of SYK in human adaptive NK cells. Some subsets of mature human T cells uniquely coexpress SYK and ZAP70.⁴¹ To address the effects of SYK coexpression in a ZAP70-bearing T cell, Syk was transduced into the prototypic Jurkat T cell line, which has been used for the identification of many ITAM-based signaling pathways (Fig. 4A).⁴² The expression of SYK was confirmed by Western blot and flow cytometric analysis (Fig. 4B, C). Coexpression of SYK did not affect the endogenous levels of expression for ZAP70, CD3ε, or CD3ζ or affect basal levels of CD69, indicative of no overt activation of the cells (data not shown). Anti-CD3 stimulation of the WT and SYK-bearing Jurkat cells resulted in diminished Ca²⁺ influx in the Jurkat cells expressing SYK (Fig. 4D, E) and significantly lower phosphorylation of LYN (Tyr397), LCK (Tyr394), HCK (Tyr411), BLK (Tyr389), ZAP70 (Tyr493), pZAP70 (Tyr292), and ERK1/2 (Thr202/Tyr204) compared with WT cells (Fig. 4F, G). Diminished activation in SYK-bearing Jurkat cells were also reflected by a delayed induction of CD69 after activation (Fig. 4H, I). Proliferation of SYK-bearing Jurkat cells that endogenously coexpressed ZAP70 significantly increased over 4 d in culture compared with WT Jurkat cells expressing only ZAP70 (Fig. 4J). Enhanced proliferation in culture was strikingly observed when SYK was expressed at 2 levels (high and low) in CD19-CAR T cells (Fig. S2). Proliferation was observed in the absence of CAR engagement indicating that expression of SYK alone caused the proliferation potentially due to tonic activity of SYK in the absence of ZAP70. The difference in the effects of SYK in primary T cells compared with Jurkat may be because Jurkat cells are a transformed cell line that constitutive proliferates.

Figure 4.

SYK expression in Jurkat T cells diminishes ITAM signaling while promoting proliferation. (A–H) Human T cell line Jurkat, constitutively expressing endogenous ZAP70, was transduced with a retroviral vector expressing SYK. (A) Diagram of WT and SYK-expressing Jurkat cells with associated biological outcomes. (B) Western blot analysis of Jurkat cells with anti-SYK, anti-ZAP70, and loading control anti-GAPDH antibodies. (C) Intracellular SYK and ZAP70 were analyzed by flow cytometry. (D) Calcium influx assay of unstimulated (left graph), and anti-CD3 crosslinked activation Jurkat cells (right graphs). (E) Summary of the calcium influx for WT and SYK-expressing Jurkat cells; the far right graph represents the calculated area under the curve (AUC) for each cell line and replicates. (F) Heatmap of phosphorylation kinetics of anti-CD3 crosslinked Jurkat cells for each signaling molecule. (G) Representative line graphs of geometric mean fluorescence intensity (gMFI) or pERK1/2 (Thr202/Ty204) from data in panel F. (H, I) Plate-bound anti-CD3 stimulated Jurkat cells analyzed for CD69 gMFI for 96 h and percentage. *P < 0.05. (J) Flow cytometric analysis of proliferation of Jurkat cells cultured with media and 200 U/mL IL2 as measured with counting beads over 96 h. The sample t test with significance is shown as *P < 0.05 and ****P < 0.0001. Data in panels E through I represent 3 to 4 biological replicates for 2 or more independent experiments. PE, phycoerythrin; UV, ultraviolet.

To identify the transcriptional effects of SYK and ZAP70 in the KHYG1 and Jurkat cells, RNA sequencing was performed on WT, $\Delta$SYK, and $\Delta$ZAP70 KHYG1 cells and on WT and SYK-bearing Jurkat cells. At the basal state without deliberate ITAM-based receptor stimulation many genes were upregulated and downregulated in the variant ZAP70 and SYK cells. Principal component (PC) analysis showed that the transcriptional profile of $\Delta$SYK KHYG1 cells was most similar to WT KHYG1 cells (PC2; 20% variance explained), while $\Delta$ZAP70 KHYG1 cells were most variable compared with the others (PC1; 64% variance explained) (Fig. 5A). More differentially expressed (DE) genes were observed in $\Delta$ZAP70 KHYG1 cells compared with $\Delta$SYK KHYG1 cells with 161 of 3,148 total DE genes overlapping (Fig. 5B). Genetic pathway analysis revealed significantly enriched pathways in “protein tyrosine kinase activity,” “regulation of MAPK cascade,” “regulation of ERK1 and ERK2 cascade,” and “calcium ion binding” in $\Delta$SYK KHYG1 cells compared with WT KHYG1 cells (Fig. 5C, left). $\Delta$ZAP70 KHYG1 cells revealed significantly enriched gene pathways in metabolism and growth compared with WT KHYG1 cells (Fig. 5C, right). Heatmap analysis of $\Delta$ZAP70 KHYG1 cells showed enrichment of genes related to phosphatase activity, pyruvate metabolic, and glucose metabolic process compared with WT and $\Delta \mathrm{SYK}$ KHYG1 cells (Fig. 5D). DE gene analysis of SYK-bearing Jurkat cells resulted in more significantly downregulated genes compared with WT Jurkat cells (Fig. 5E). Pathway analysis revealed significantly suppressed pathways in calcium-dependent “phospholipid binding,” “MAP kinase kinase activity,” “regulation of receptor signaling pathway via STAT,” and “activation of protein kinase activity” in SYK-bearing Jurkat cells compared with WT Jurkat cells (Fig. 5F). Together, these data support a negative role for SYK in regard to phosphorylation and transcriptional programing in NK and T cells, whereas ZAP70 is essential for these functions.

Figure 5.

RNA sequencing of SYK and ZAP70 variant cell lines. (A-F) RNA sequencing of cells. (A) PC analysis of normalized KHYG1 gene expression and percent variants on the axis. (B) Venn diagram of overlapping significant differential gene expression comparing WT KHYG1 with KHYG1 cells ablated of SYK or ZAP70. Top genes with significance of an adjusted P value of <0.05 were only considered in the Venn diagram. (C) Top-ranked gene set enrichment analysis of the top 10 to 11 chosen activated pathways enriched in KHYG1 cells depleted of SYK or ZAP70. (D) Heatmap of significant genes from the top pathways observed in panel C for each KHYG1 cell type. (E) DE gene expression between WT and SYK-transduced Jurkat cells. The top significant genes were plotted, and orange dots represent significant genes; blue dots represent nonsignificant genes. (F) Top-ranked gene set enrichment analysis of the top 9 chosen suppressed pathways observed in Syk-transduced Jurkat cells. Data used for each figure consisted of 3 biological replicates for each cell line.

Ablation of SYK in primary human CAR NK cells enhances tumor cell killing

The enhanced ITAM signaling in $\Delta$SYK KHYG1 cells strongly suggest that ablation of SYK in primary human NK cells might enhance the function of CAR NK cells. Primary NK cells were transduced with transduced with a CD19-targeting CAR containing cytoplasmic signaling elements of CD28 and CD3ζ, and then CRISPR editing was used to ablate ZAP70 or SYK (Fig. 6A–C). We first confirmed by flow cytometry that genetic depletion of SYK or ZAP70 did not alter surface expression of CD16, CD56, NKp30, NKp44, or NKp46 (data not shown). To test the influence of SYK and ZAP70, we conducted in vitro cytotoxicity assays with WT CAR NK cells, $\Delta$ZAP70 CAR NK cells, and $\Delta$SYK CAR NK cells against SUPB15, a CD19-expressing B lymphoblast cell line, for 4 and 24 h coculture (Fig. 6D, E). SYK-ablated CAR NK cells mediated the most potent tumor-specific killing at 4 and 24 h, whereas ZAP70-ablated CAR NK cells mediated the lowest tumor-specific killing compared with WT NK cells (Fig. 6D, E). Similarly, CAR NK cells ablated of SYK also showed enhanced cytokine production (tumor necrosis factor α and interferon $\mathrm{\gamma }$) and degranulation as measured by surface CD107a compared with WT CAR NK cells, whereas $\Delta$ZAP70 CAR NK cells had significantly diminished effector functions (Fig. 6F–I). Overall, our study shows that the ablation of SYK in NK cells enhanced CAR NK cell–mediated killing and increased production of effector cytokines against tumor targets (Fig. 6A).

Figure 6.

SYK ablation enhances CD19-CAR function in human primary NK cells. (A–I) SYK-ablated primary human NK cells demonstrated enhanced killing and cytokine responses against the SUPB15 target cell line. (A) Depiction of ZAP70 signaling driving enhanced target cell killing, degranulation, and cytokine production in primary CAR NK cells. (B) Diagram of the 9-d introduction of the retroviral 1928ζ CAR construct into primary NK cells and subsequent depletion of SYK or ZAP70 using CRISPR-Cas9. (C) Flow cytometry–based validation of CAR expression and SYK or ZAP70 depletion in primary NK cells, gated on live CD3⁻CD14⁻CD19⁻CD56⁺ NK cells, nontransduced control (NT). (D) Summary graph of the 4-h flow cytometry–based killing assay. (E) Summary graph of the 24-h luciferase-based killing assay. (F–I) Markers identified after 4-h coculture with SUPB15 by flow cytometry. (F) Percentage of CD107a, interferon γ (IFNγ), and tumor necrosis factor α (TNFα); plots of cells gated on live CD19-CAR⁺ NK cells. (G) Summary graph from data in panel F. (H) Flow cytometry plot of CD107a frequency, gated on live CD56⁺ CAR⁺ NK cells. (I) Summary plot of data shown in panel H. Data used in this figure consisted of 3 technical and 2 biological replicates per figure. The sample t test with significance is shown as *P<0.05, ***P<0.001, and ****P <0.0001. FITC, Fluorescein Isothiocyanate; PerCP, Peridinin-chlorophyll-protein; BUV, Brilliant Ultra Violet; AF647, Alexa Fluor 647; PE, phycoerythrin; scFV, single-chain variable fragment.

Discussion

SYK and ZAP70 tyrosine kinases are present in all jawed vertebrates and are involved in immune receptor functions.^11,12 They contain SH2-binding and kinase domains that mediate their activity. SYK is more broadly expressed, being present in myeloid cells, B cells, T cells, and NK cells, whereas ZAP70 is predominantly expressed in T cells and NK cells. Mouse immature “double-negative” T cells in the thymus coexpress both SYK and ZAP70, but typically lose SYK expression during maturation, although SYK is present in some γδ T cells, mucosal-associated invariant T cells, and invariant NK T cells (www.immgen.org). Mice lacking Zap70 are blocked at the “double-positive” stage of development and are largely devoid of mature αβ TCR T cells in peripheral lymphoid organs.^11,43 SYK is involved in pre-TCR signaling in double negative thymocytes in mice and lack of Syk results in impaired T cell maturation.⁴³ In humans, immature T cells in the thymus coexpress SYK and ZAP70; however, unlike in mice, humans with loss-of-function ZAP70 genes lack CD8⁺ αβ TCR T cells in the periphery but have abundant CD4⁺ αβ TCR T cells due to expression and redundant function of SYK, albeit with defective function.⁴⁴

Mouse NK cells in all stages of development coexpress SYK and ZAP70, but interestingly ILC2 and ILC3 lack SYK but express ZAP70 (www.immgen.org). Normal numbers of NK cells are present in bone marrow chimeric Rag2^−/−Il2rg^−/− mice reconstituted with bone marrow with disrupted Syk genes and found that NK cell development was largely normal, potentially due to the redundant function of Zap70.⁴⁵ Natural cytotoxic activity was intact in Syk^−/− mice, although they demonstrated slightly reduced ADCC activity.⁴⁵ Mixed bone marrow chimeric Rag2^−/−Il2rg^−/− mice reconstituted with Syk^−/− fetal liver cells demonstrated normal numbers of NK cells and natural cytotoxicity was intact, whereas B and T cells were absent.⁴⁶ However, induction of IFNγ secretion by stimulation with agonist antibodies to CD16 (signaling through FcεR1γ) and Ly49D (signaling through DAP12) were absent, although surprisingly stimulation through NK1.1 was retained, implying an SYK and ZAP70-independent mechanism.⁴⁶

After clonal expansion induced by CMV infection, human NK cells can acquire an adaptive or memory state characterized by the loss of SYK and FcεR1γ expression, accompanied by enhanced ADCC function.^3,4,7 Human adaptive NK cells rely on CD3ζ and ZAP70 to activate these NK cells when initiated by the CD16 Fc receptor, similar to T cells using CD3ζ and ZAP70 to transmit TCR-induced signals. Loss of SYK and FcεR1γ does not occur in adaptive or memory mouse NK cells driven by mouse CMV infection. Unlike human CD3ζ, mouse CD3ζ cannot associate with CD16, so loss of FcεR1γ in mice results in total loss of ADCC function.¹⁶ Humans with loss-of-function ZAP70 mutations have normal numbers of NK cells.⁴⁴

Given the loss of SYK in human adaptive NK cells, we investigated the roles of SYK and ZAP70 on downstream signal transduction in a model system ablating SYK or ZAP70 in a clonal human NK cell line. Our study confirms and extends the recent findings of Dahlvang et al.,¹³ which demonstrated enhanced cytotoxicity in NK cells by ablation of SYK. Using a mechanistic in silico model, we showed that the competition between ZAP70 and SYK in accessing a limited number of ITAMs leads to decrease in Ca²⁺ flux in WT compared with the ΔSYK KHYG1 NK cells. Our model predicts that the presence of a higher number of ITAMs reduces the competition between ZAP70 and SYK for accessing phosphorylated ITAMs and thus can produce higher Ca²⁺ flux and potentially increased cytotoxicity in WT compared with ΔSYK NK cells. Our studies reveal that whereas SYK preferentially promotes proliferation, ZAP70 enhances Ca²⁺ influx and downstream activation when ITAM-based receptors are engaged. Similar results were obtained when Jurkat T cells endogenously expressing ZAP70 were transduced with SYK; proliferation was enhanced but TCR-induced activation was diminished. Expression of SYK in primary CAR T cells also dramatically enhanced proliferation, and induction of proliferation did not require engagement of the CAR. These findings suggest that tonic stimulation by SYK, which is a more active kinase than ZAP70 and less dependent on Src family kinases for activation,⁴⁷ may be responsible for this elevated proliferation.

Analysis of KHYG1 NK cells and Jurkat cells expressing ZAP70 and/or SYK revealed extensive transcriptional alterations impacting the differential behavior of these cells. The more robust effector functions of the KHYG1 and Jurkat cells lacking SYK and expressing only ZAP70 correlated with more active kinase, Ca⁺⁺ signaling, and ERK signaling pathways, perhaps accounting for the more robust ADCC function in human adaptive NK cells lacking SYK. The elevated proliferation in the KHYG1 NK, Jurkat, and CAR T cells expressing SYK may reflect the enhanced expression of genes in the metabolic pathways promoting proliferation. Ablation of ZAP70 or SYK in primary NK cells expressing a CD19-targeted CAR confirmed superior effector function in NK cell lacking SYK using in vitro assays to measure cytotoxicity and cytokine production. Further studies will be needed to evaluate the function of these SYK-ablated CAR NK cells in vivo and effects on proliferation and durability. Collectively, these studies reveal differential roles for SYK and ZAP70 in T and NK cells with regard to ITAM-mediated signaling, proliferation, and effector function.

Author contributions

A.J.M. and O.A.A. contributed to conception, design, data collection, analysis and interpretation, and writing and editing; V.A. and I.N. contributed to conception, design, data collection, analysis and interpretation, and writing; J.B.L., L.M.Q.-M., J.S.A.-H., G.U., A.G.R., A.S. contributed to data collection, J.E., J.D., and L.L.L. contributed to conception, design, data analysis and interpretation, and writing and editing.

Supplementary material

Supplementary material is available at The Journal of Immunology online.

Funding

This work was supported by National Institutes of Health grant R01AI146581 and the Parker Institute for Cancer Immunotherapy. Virtual Cell is supported by National Institutes of Health grant R24 GM137787 from the National Institute for General Medical Sciences.

Conflicts of interest

L.L.L. has served on the advisory board of Cullinan Oncology, Dragonfly, DrenBio, Edity, GV20, IMIDomics, InnDura Therapeutics, Innovent, Mendus, Nextpoint, Nkarta, oNKo, Obsidian Therapeutics, Stamford Pharma, and SBI Biotech to advance NK cell–based therapies.