Graphical abstract graphic file with name fx1.jpg [47]Open in a new tab Highlights * • NK and T cells produce TNF differently due to stimulation thresholds and TNF trafficking * • Engaging activating receptors and providing IL-2 enhances the effectiveness of NKCEs * • NKCE, TCE, and IL-2 therapies exhibit both shared and different therapeutic advantages __________________________________________________________________ Hang et al. compare NK cell engagers (NKCEs) and T cell engagers (TCEs), revealing intrinsic differences in the responses of cells upon engagement. This study highlights design strategies to improve the efficacy of NKCEs while minimizing the risk of cytokine release syndrome, which is often associated with T cell-targeted therapies. Introduction T cell engagers (TCEs) are targeted immunotherapies that redirect endogenous T cells into proximity of cancer cells with a tumor targeting domain and trigger T cell signaling via a CD3 crosslinking domain.[48]^1 The therapeutic value of this approach has been demonstrated by the clinical approval of blinatumomab, tebentafusp, and teclistamab.[49]^2^,[50]^3^,[51]^4 However, widespread use of TCE treatments is limited by toxicity due to the induction of cytokine release syndrome (CRS) upon T cell receptor (TCR) engagement.[52]^5 Although efficacious, TCE liabilities have prompted the investigation of alternative immune cell types that mediate cytotoxic effects without causing CRS. Like cytolytic T cells, natural killer (NK) cells can kill tumor cells through the targeted release of perforin and granzymes from lytic granules.[53]^6 Despite similar degranulation tendencies after activation, T cells are better poised to produce tumor necrosis factor (TNF),[54]^7 which is believed to trigger monocyte activation and adverse events linked to CRS during TCE administration.[55]^8 TNF biosynthesis is tightly regulated at the transcriptional and posttranscriptional levels.[56]^9 Initially synthesized as a transmembrane protein, TNF requires proteolytic cleavage by TNF-converting enzyme (TACE/ADAM17) to become its soluble form.[57]^10 The differential regulation of cytokines, including TNF, by T cells and NK cells remains unclear. Unlike T cells, NK cells recognize tumor cells through a set of stimulatory and inhibitory receptors. NK cells express activating receptors such as CD16 (FcγRIIIA), natural killer group 2 member D (NKG2D), and natural cytotoxicity triggering receptor 1 (NCR1/NKp46).[58]^11 Both NCR1 and CD16 associate with immune receptor tyrosine-based activation motif (ITAM)-containing CD3ζ and FcεRγ to trigger downstream signaling, while NKG2D signals through the adaptor DNAX-activation protein 12 (DAP12).[59]^12 By binding to the fragment crystallizable (Fc) region of immunoglobulin Gs (IgGs), CD16 engagement allows NK cells to initiate antibody-dependent cellular cytotoxicity (ADCC) against antibody-coated cancer cells. Furthermore, afucosylation of Fc increases the affinity of Fc to CD16, leading to enhanced ADCC activity.[60]^13 NCR1 has been identified as a promising target for the development of NK cell engagers (NKCEs) because of its sustained expression on tumor-infiltrating NK cells and potential strategy to engage both NCR1 and CD16 through incorporation of an Fc domain.[61]^14 Despite their potent antitumor activity, the development of NKCEs faces several major challenges. First, many tumors are sparsely infiltrated with NK cells.[62]^6 Second, NK cells exhibit low proliferative capabilities.[63]^6 Third, NK cells downregulate stimulatory receptors such as CD16 upon activation, resulting in a lack of sustained antitumor response.[64]^11 To overcome these obstacles, we designed tumor-targeted NKCEs that incorporate interleukin (IL)-2 to sustain NK proliferation, survival, and function and engage both NCR1 and CD16 simultaneously to bypass reduced receptor expression.[65]^11 Through benchmarking experiments, we identified optimal NKCE formats that induced sustained killing against tumor cells with minimal cytokine production, thus representing a next generation of molecules for cancer treatment. Results NKCEs redirect NK cells to kill target cells with minimal cytokine release To compare NKCEs with TCEs side by side, we initiated our studies by generating a bispecific NKCE tool molecule by linking a single-chain variable fragment (scFv) targeting the common tumor antigen HER2, with an scFv targeting the NK cell receptor NCR1 ([66]Figure 1A). We designated this bispecific molecule as NCR1xHER2 NKCE. A TCE tool molecule (CD3xHER2) was generated in analogous fashion to the NKCE tool molecule by linking a scFv targeting CD3 to the anti-HER2 scFv ([67]Figure 1A). The binding affinities as represented by the dissociation constants (K[D]) of NCR1xHER2 and CD3xHER2 for their respective targets was 1.0 (HER2), 10.3 (NCR1), and 3.7 (CD3) nM, indicating similar affinities for NCR1 and CD3 ([68]Table S1). JIMT1 tumor cells expressing endogenous HER2 and firefly luciferase (JIMT1-Luc) were employed as target cells for evaluating tool molecule mediated tumor killing, while HER2 knockout cells (HER2^−/− JIMT1-Luc) served as a negative control ([69]Figure 1B). To compare the effects of NKCE and TCE on tumors, peripheral blood mononuclear cell (PBMC)-derived NK or Pan T cells were mixed with engager molecules, and tumor cell lysis was measured by quantifying luciferase activity. 24 h post-incubation, both NKCE and TCE molecules demonstrated antigen-specific responses by inducing lysis of HER2-positive tumor cells, while having no effect on HER2^−/− cells ([70]Figure 1C). In addition, a comparison of killing activity at 96 and 168 h revealed that the TCE resulted in a higher maximum killing (Emax) and sustained killing effect on tumor cells compared to the NKCE ([71]Figures S1A and S1B). Figure 1. [72]Figure 1 [73]Open in a new tab NKCEs redirect NK cells to kill target cells with minimal cytokine release (A) Design of tool molecules NCR1xHER2 (NKCE) and CD3xHER2 (TCE). (B) HER2 expression on JIMT1-Luc WT and HER2^−/−. Antibodies bound per cell (ABCs) of HER2 are indicated in the graph. (C and D) JIMT1-Luc WT and HER2^−/− were incubated with NK cells or Pan T cells at an E:T cell ratio of 5:1 and increasing concentrations of NCR1xHER2 or CD3xHER2. (C) JIMT1-Luc cell lysis at 24 h; concentrations of TNF and IFNγ (D) in the cell culture supernatants at 24 h. (E) Design of tool molecules NCR1xCLL1 (NKCE-I) and CD3xCLL1 (huCLL1 TCE). (F) CLL1 expression on PL21-Luc, NOMO1-Luc, and MOLM13-Luc. ABC of CLL1 is indicated in the graph. (G and H) PL21-Luc, NOMO1-Luc, and MOLM13-Luc were incubated with NK cells or Pan T cells at an E:T cell ratio of 2.5:1 and increasing concentrations of NKCE-I or huCLL1 TCE. (G) Target cell lysis at 48 h; (H) concentrations of TNF in the cell culture supernatants at 48 h. Data are presented as means ± SEM (C). The data shown are representative of at least two independent experiments. Next, we examined the effects of engagers on cytokine release upon tumor cell recognition. Incubation of CD3xHER2 TCE with Pan T cells showed robust TNF and interferon (IFN)γ release upon incubation with JIMT1-Luc cells ([74]Figure 1D). Both CD4^+ T and CD8^+ T cells induced tumor lysis effectively ([75]Figure S1C), but CD4^+ T cells produced more TNF and IL-2 relative to CD8^+ T cells in the presence of CD3xHER2 ([76]Figures S1D and S1E). However, IFNγ release was similar between T cell subsets upon TCE engagement ([77]Figure S1F). This indicates different regulation of cytokines within T cell populations, which aligns with recent reports on cytokine production by CD4^+ and CD8^+ chimeric antigen receptor (CAR)-T cells.[78]^15^,[79]^16 In stark contrast to T cells results, NK cells incubated with tumor cells and NCR1xHER2 showed no discernable release of either TNF, IFNγ, or IL-2 but retained tumor-specific lysis ([80]Figures 1C, 1D, and [81]S1E). Thus, unlike T cells, NK cell-mediated tumor lysis may be uncoupled from cytokine release. To further compare NKCEs and TCEs, we investigated the effects of tumor antigen expression levels on cytokine release and tumor lysis upon engagement. We began by creating tool engagers against C-type lectin-like molecule-1 (CLL1), a tumor antigen expressed at varying degrees in acute myeloid leukemia cells.[82]^17 An NKCE against CLL1 (NCR1xCLL1), referred to hereafter as NKCE-I, was developed on a human IgG1 stable effector functionless (SEFL)2 backbone to abrogate FcγR binding.[83]^18 A half-life extended (HLE) huCLL1 TCE consisting of an Fc domain served as a comparative TCE control ([84]Figure 1E).[85]^19 Affinity measurement revealed that NKCE-I and huCLL1 TCE bound NCR1, CD3, and huCLL1 with K[D] ranging from 4.9 to 20 nM ([86]Table S1). To evaluate engagers, three cell lines with different levels of surface CCL1 expression (PL21, NOMO1, and MOLM13) were used to express luciferase (Luc) ([87]Figure 1F). In tumor killing assays, NKCE-I showed an Emax of 76% tumor lysis in PL21-Luc (CCL1 high), an Emax of 21% in NOMO1-Luc cells (CLL1 medium), and undetectable killing in MOLM13-Luc (CLL1 low) ([88]Figure 1G). In contrast, huCLL1 TCE was able to achieve at least 70% Emax against all three cell lines tested regardless of antigen expression with half maximal effective concentration (EC50) in the pM range ([89]Figure 1G). Furthermore, like results seen with HER2 engagers, huCLL1 TCE, but not NKCE-I, induced robust TNF release in killing assays against CLL1-expressing cells ([90]Figure 1H). Altogether, comparative studies between analyzed TCE and NKCE tool molecules revealed that NKCEs could direct NK cells to kill tumor cells with minimal cytokine release. However, the efficacy of NKCE killing is more dependent on target antigen density than TCEs. Differential TNF production due to stimulation threshold and TNF trafficking Minimal TNF release induced by NKCEs makes them an attractive alternative to TCEs, which prompted us to investigate whether differential cytokine production by NK and T cells is solely dependent on the degree of activation receptor engagement or an inherent cell intrinsic property. We first tested if direct engagement of their respective activation receptors (NCR1, CD16, and CD3/TCR) in the absence of target cells or other interactions resulted in differential TNF production. Receptor density quantification showed that CD3 expression on Pan T cells is 12-fold of NCR1 expression and twice as high as CD16 expression on NK cells ([91]Figure 2A). For analysis of activation receptor-mediated cytokine production, NK cells or T cells were stimulated with immobilized anti-NCR1, anti-CD16, or anti-CD3 antibodies on plates, and total intracellular TNF levels were measured via western blot (WB) analysis. This was conducted in the presence of a protein transport inhibitor (GolgiPlug) that prevents TNF secretion and results in the accumulation of the immature 25 kDa TNF form.[92]^10 Although variation in intracellular TNF precursor levels was observed among donors, T cells produced significantly higher TNF protein following anti-CD3 stimulation, compared to NK cells stimulated with anti-NCR1 or anti-CD16 ([93]Figures 2B and [94]S2A). Figure 2. [95]Figure 2 [96]Open in a new tab Differential TNF production due to stimulation thresholds and TNF trafficking (A) CD3 expression on Pan T cells and NCR1 or CD16 expression on NK cells from 6 donors were measured by flow staining to determine ABC. (B) Immunoblot of lysates of Pan T cells or NK cells from 6 donors stimulated with immobilized anti-CD3 or anti-NCR1 and anti-CD16, respectively, for 16 h in the presence of GolgiPlug. The ratio of TNF/actin from each donor was calculated. (C) Immunoblot of lysates of Pan T cells or NK cells from 6 donors stimulated with PMA/ionomycin in the presence of transporter inhibitors (Ti). The ratio of TNF/actin from each donor was calculated. (D) TNF in the cell culture supernatants of Pan T cells or NK cells from 8 different donors at 4 h post-stimulation. (E–G) Surface staining of TNF on Pan T or NK cells from 6 donors in the presence of DMSO or GW280264X post-4 h PMA/ionomycin stimulation. (E) Representative staining, (F) frequency of TNF^+ among Pan T or NK cells, and (G) mean fluorescence intensity (MFI) of TNF^+ among Pan T or NK cells. (H) Immunoblot of lysates of Pan T cells or NK cells from 7 donors analyzed with anti-ADAM17 and anti-actin. The ratio of ADAM17/actin from each donor was calculated. Data are presented as means ± SEM (A). p values were determined by one-way ANOVA followed by Tukey’s multiple comparisons test (A and B) or a paired two-tailed t test for two-group comparisons (C, D, F, G, and H); ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. To further define whether difference in TNF expression between NK and T cells is dependent on the strength of antibody-mediated activation of receptors or is a cell-intrinsic property, we analyzed cytokine production using a small molecule cocktail of phorbol 12-myristate 13-acetate (PMA) and ionomycin. PMA/ionomycin directly activates nuclear factor of activated T cells (NFAT) and protein kinase C intracellular signaling pathways to induce cytokine production independent of the activation receptors agonized by engager, thus serving as a control for maximal cellular activation.[97]^20 Additionally, secreted and intracellular TNF were both measured to determine if any observed differences depended on secretion or production of TNF protein. Under PMA/ionomycin stimulation, similar levels of intracellular TNF precursor protein were produced in NK and T cells, suggesting that both cell types have equal capacity to produce TNF protein under equivalent levels of maximum stimulation ([98]Figures 2C and [99]S2B). In contrast, under PMA/ionomycin stimulation, T cells secreted significantly higher levels of TNF relative to NK cells, suggesting intrinsic differences between cell types ([100]Figure 2D). The discrepancy between intracellular and secreted TNF suggests that these inherent differences may lie in their capability to process and secrete TNF. TNF exists initially as a cell surface transmembrane protein, requiring proteolytic cleavage by ADAM17 to be processed into its soluble form.[101]^10 To determine whether differences in membrane levels of TNF or ADAM17-mediated cleavage contribute to differential TNF secretion between cell types, we performed PMA/ionomycin stimulation experiments using the potent ADAM17 inhibitor GW280264X to allow detection of surface membrane-bound TNF.[102]^21 In the presence of ADAM17 inhibition, NK cells had significantly less surface expression of TNF relative to T cells upon PMA/ionomycin stimulation, suggesting that trafficking of TNF precursors to the plasma membrane may account for differences between NK cells and T cells ([103]Figures 2E–2G). Additionally, WB analysis revealed lower protein expression of ADAM17 in NK cells relative to T cells ([104]Figures 2H and [105]S2C), suggesting that differences in TNF processing by virtue of reduced ADAM17 expression may also contribute to differences in secretion between the two cell types. Altogether, these data suggest that innate differences in both the ability of TNF to be trafficked to the plasma membrane and the ability to be processed to its soluble form may contribute to the inherent TNF secretion differences between NK cells and T cells. Multifunctional NKCE fusion with IL-2 enables sustained NK cell-mediated cytotoxicity of tumor cells Although the evaluated tool NKCE molecules induce less cytokine production upon tumor cell engagement, they lack sustained killing of tumor cells compared to tool TCEs ([106]Figures S1A and S1B). TCR engagement via TCEs leads to the release of IL-2 ([107]Figure S1E), which sustains T cell proliferation and allows for enhanced tumor killing.[108]^22^,[109]^23 Therefore, we hypothesized that an NKCE combined with IL-2 could similarly sustain NK cell survival and enhance lysis of NOMO1-Luc cells, which express less target antigen and are less sensitive to NKCE-mediated killing ([110]Figures 1F and 1G). Indeed, the combination of NKCE-I with IL-2 led to 100% Emax killing of NOMO1-Luc cells, whereas NKCE-I or IL-2 alone resulted in respective levels of 37% and 81% Emax ([111]Figure S3A). The addition of IL-2 also resulted in a dose-dependent increase in live NK cells and reduced tumor viability in treated groups after 96 h ([112]Figures S3B and S3C). Cell tracer experiments further demonstrated that IL-2 supported NK viability by increasing proliferation in IL-2-treated groups ([113]Figure S3D). Analysis of T cells treated with TCE showed robust IL-2 production during killing assay, demonstrating their intrinsic ability to support their own proliferation by producing IL-2 ([114]Figures S3D and S3E). Thus, combining IL-2 with NKCE represents an attractive strategy to boost NKCE activity against tumor cells with low antigen expression, which require sustained killing. To further identify ways to improve NKCEs, we investigated whether engaging a second activating receptor (CD16) could enhance tumor killing alone and synergize with the addition of IL-2. To test this, we developed several NKCE molecules targeting CLL1 that contained IL-2 and variants of IgG1 that do not engage (SEFL2 Fc) or engage (afucosylated Fc) CD16 ([115]Figure 3A). NKCE-I is used as a benchmark for solely engaging one activating receptor (NCR1), while NKCE-II was created to engage both NCR1 and CD16 on NK cells by using afucosylated Fc ([116]Figure 3A). Because NKCE-II engages two co-receptors and has the potential to bring separate NK cells together in trans, we began by determining if NKCE molecules could induce NK cell fratricide. As expected, NK cells incubated with NKCE-I resulted in no evidence of fratricide in a 96 h assays, while NKCE-II induced mild fratricide evident only at high concentrations (EC50: 643 pM) and was unable to achieve Emax above 30% ([117]Figures S4A and S4B). In tumor co-culture assays, both NKCE-I and NKCE-II exhibited potent cytotoxic activity against PL21-Luc ([118]Figure S4C). NKCE-II demonstrated superior cytotoxic activity relative to NKCE-I, with a >10-fold lower EC50 value and a more pronounced dose-dependent reduction in tumor cells ([119]Figures S4C and S4D). Altogether, these studies demonstrated that co-engagement of two activating receptors (CD16 and NCR1) achieves superior tumor killing than engagement of a single receptor (NCR1), with minimal fratricidal effects well below the effective concentration achieved for killing tumor cells. Figure 3. [120]Figure 3 [121]Open in a new tab Multiple functional NKCEs induce robust killing activity with minimal TNF production (A) Components and designs of multifunctional NKCEs: NKCE-I to NKCE-V. (B–D) NOMO1-Luc cells were incubated with NK cells or Pan T cells at an E:T cell ratio of 2.5:1 and increasing concentrations of NKCE-I to NKCE-V and IL-2 or TCE. (B) NOMO1-Luc lysis at 48 h. EC50 and area under curve (AUC) are indicated in the graph. Concentrations of TNF (C) and IFNγ (D) in the cell culture supernatants at 48 h. (E) CLL1 expression on PL21-Luc WT and CLL1^low. ABC of CLL1 is indicated in the graph. (F–H) PL21-Luc WT or CLL1^low cells were incubated with NK cells at an E:T cell ratio of 2.5:1 and increasing concentrations of NKCE-III (F), NKCE-IV (G), and NKCE-V (H) and PL21-Luc lysis at 48 h. Data are presented as means ± SEM (B). The data shown are representative of at least two independent experiments. Next we evaluated the optimal IL-2 format to synergize with NCR1 or CD16 engagement and generated three additional NKCE molecules. NKCE-III was created to deliver IL-2 and engage only NCR1, while NKCE-IV was created to deliver IL-2 and solely engage CD16 ([122]Figure 3A). Lastly, NKCE-V was created to deliver IL-2 and engage both NCR1 and CD16 ([123]Figure 3A). For NKCEs that engaged NCR1, binding affinities for NCR1 were relatively similar and K[D] ranged from 15 to 20 nM ([124]Table S2). The binding affinity with CD16a varies significantly: NKCE-I and NKCE-III show no significant binding, as expected, while NKCE-II, NKCE-IV, and NKCE-V demonstrate binding affinities ranging from 0.6–1 nM (CD16a158V) to 3.6–5.1 nM (CD16a158F) ([125]Table S2). We characterized these multifunctional NKCEs in a killing assay against NOMO1-Luc tumor cells while using huCLL1 TCE as a control. All NKCEs efficiently induce NK cell-mediated killing of NOMO1-Luc cells but with varying potency: NKCE-V > NKCE-IV > NKCE-III > NKCE-II > NKCE-I ([126]Figure 3B). The potency in the killing assay reflects differences in the type(s) of receptor engagement and addition of IL-2 ([127]Figure 3A). While NKCE-I and NKCE-II induced comparable Emax level of killing against NOMO1-Luc, NKCE-II showed a lower EC50 relative to NKCE-I, like results seen with PL21-Luc ([128]Figure S4C). NKCE-III, which engages one activating receptor (NCR1), plus IL-2R, outperformed its respective counterpart that only engaged NCR1 (NKCE-I), resembling T cell-mediated killing induced by the control TCE molecule. Moreover, NKCE-V, which engages both co-receptors plus IL-2R, outperformed all NKCE molecules, demonstrating synergism of all three pathways. All NKCEs induced low TNF production, with a maximum of 50 pg/mL at 48 h ([129]Figure 3C), while IFNγ production induced by different NKCEs was in line with their respective potencies, with the highest IFNγ detected in the NKCE-V treatment ([130]Figure 3D). However, regardless of NKCE format, NK cells produced much lower TNF and IFNγ relative to T cells upon engagement. Lastly, we tested whether multifunctional NKCEs could maintain tumor killing under low tumor antigen levels by comparing NK cell-mediated killing against PL21-Luc cells expressing wild type (WT) and an isogenic CLL1 low-expressing cell line (CLL1^low) ([131]Figure 3E). NKCE-III, NKCE-IV, and NKCE-V demonstrated killing of PL21-Luc WT cells but displayed significantly less potency and Emax against CLL1^low cells ([132]Figures 3F–3H). However, NKCE-V showed the most robust killing against CLL1^low cells among all NKCEs tested, indicating that engaging both co-receptors plus IL-2R may reduce the barrier of killing of low-antigen tumors. Altogether, these data demonstrate that engaging multiple co-activating receptors plus delivering IL-2 is an optimal strategy to enhance NKCE activity against tumor cells. Transcriptomic landscape of T cell subsets and NK cells during antitumor response Given that IL-2 addition to NKCEs led to sustained tumor cell killing and improved NK viability ([133]Figures 3B and [134]S3), we investigated how engagement of IL-2R enhances NK function in comparison to TCE-treated T cells. We performed transcriptome profiling of NK cells and T cell subsets (CD4 and CD8) during antitumor responses using various engagers ([135]Figure 4A). To normalize for target engagement, we used PL21-Luc as the target cells due to the similar killing efficacy between NKCEs and control TCE ([136]Figure S5A). Transcriptomic changes were analyzed at 16 and 64 h post-treatment to identify gene signatures linked to sustained activity. Although decreasing over time, sorted NK cell numbers were comparable between groups within 16 and 64 h time points ([137]Figure S5B). Principal-component analysis (PCA) revealed distinct clustering by cell type, which represents the most variance (PC1), confirming expected transcriptional differences between T cell subsets and NK cells ([138]Figure 4B). Additional variance was observed based on time point and treatment conditions (e.g., engager or PBS). Among the NK cell treatments, NKCEs incorporating IL-2 or combined with exogenous IL-2 clustered together and were separated from non-IL-2R engaging NKCE-II treatment ([139]Figures 4B and [140]S5C). Thus, the addition of IL-2 to NKCEs had the most pronounced effect on NK cell gene expression profiles. Figure 4. [141]Figure 4 [142]Open in a new tab Transcriptomic landscape of T cell subsets and NK cells during antitumor responses (A) NKCEs and the TCE used in RNA sequencing. (B–E) PL21-Luc cells were incubated with NK cells or Pan T cells from 3 donors at an E:T cell ratio of 2.5:1 with 10 nM NKCEs, IL-2, or huCLL1 TCE. (B and C) RNA sequencing of NK, CD4^+ T, and CD8^+ T cells 16 and 64 h post-different treatments. (B) PCA of data from all samples at 16 (filled color) and 64 h (clear color); (C) heatmap with hierarchical clustering comparing NK, CD4^+, and CD8^+ T cells with different treatments at 16 and 64 h. (D and E) CD4^+ and CD8^+ T cells (D) or NK cells (E) were labeled with CellTrace before the killing assay. Representative histogram graph of NKG7, Granzyme B, CellTrace, NCR1, and CD16 staining of gated NK and T cells at 96 h. To identify gene response patterns across treatments, cell types, and time points, we performed unbiased hierarchical clustering on differentially expressed genes (DEGs) with an absolute log2 fold change ≥ 2. This analysis identified 8 different clusters ([143]Figure 4C). These clusters were largely a reflection of differences between cell types (cluster 1 and 8), engager molecule potency (clusters 2 and 4), and temporal changes over time (cluster 5). Reactome pathway analysis revealed that each cluster is linked to multiple overlapping pathways ([144]Figure S5D).[145]^24 We next focused on genes affecting immune cell function across various clusters ([146]Figure 4C). Both TCE and NKCEs with IL-2 enhanced gene expression related to activation and cytotoxicity, especially in clusters 2 (TNFSF4, IL12RB2, and IFNG) and cluster 4 (LTA, LTB, and IL-2RA). At 64 h, these cells prominently increased expression of genes for cell cycling and proliferation (cluster 5: PLK2 and MKI67). NKCE-II did not induce genes in the cell cycling pathway. TCE treatment uniquely regulated costimulatory and checkpoint signaling genes (cluster 8: PDCD1 and CD28), while NKCEs influenced genes involved in degranulation and immunoregulatory interactions (cluster 1: XCL2 and SIGLEC7). Flow cytometry analysis was performed to confirm the effect of engager molecules on cytotoxicity (NKG7 and Granzyme B), activating receptors (CD16 and NCR1), checkpoints (PD-1 and T-cell immunoreceptor with Ig and ITIM domains (TIGIT)), and IL-2 signaling (CD25, gene: IL-2RA) at the protein level. Both TCE and NKCEs with IL-2 increased Granzyme B and NKG7 expression ([147]Figures 4D and 4E). This increase was linked to significant proliferation in T cells, but only mild proliferation in NK cells ([148]Figures 4D and 4E). Treatment with molecules that engage NCR1 (NKCE-II and NKCE-III) resulted in decreased NCR1 staining on NK cells, while treatment with NKCE-IV, which does not engage NCR1, led to increased NCR1 expression ([149]Figure 4E). Furthermore, all NKCE treatments resulted in CD16 upregulation on NK cells ([150]Figure 4E). Interestingly, analysis of PD-1 and TIGIT showed distinct differences by cell type. TCEs resulted in upregulation of surface PD-1 and TIGIT in T cells ([151]Figure S5E). In contrast, all NKCEs induced significant TIGIT expression but minimal PD-1 expression on NK cells at 24 h, regardless of format ([152]Figure S5F). Lastly, despite both TCE and NKCEs upregulating IL-2RA mRNA ([153]Figure 4C), only T cells exhibited an increase in CD25 expression upon TCE treatment. In contrast, CD25 levels in NK cells remained unchanged under conditions examined ([154]Figures S5E and S5F). In summary, transcriptomic analyses demonstrated both distinct and shared cellular pathways activated by NK and T cells with engager molecules. Multifunctional NKCEs promote antitumor immunity without notable adverse events Given the enhanced tumor cell killing properties seen with NKCEs incorporating IL-2, we compared the antitumor activity of NKCEs versus TCEs in vivo using immunocompromised NOD/SCID/IL2Rγ^null (NSG) mice bearing NOMO1-Luc xenograft tumors ([155]Figure 5A). NSG mice were implanted with a mixture of NOMO1-Luc and human Pan T or NK cells and given respective engager molecules. Mice treated with NKCE-III, NKCE-IV, and NKCE-V showed delayed tumor growth compared to those treated with the isotype control, with NKCE-V treatment achieving the most robust tumor growth inhibition (TGI: 83%) among tested NKCEs ([156]Figure 5B). However, xenografted mice given human Pan T cells displayed significant tumor regression in NSG mice, even without TCE treatment ([157]Figure 5B). This effect is likely due to the infused human Pan T cells attacking both the tumor and the host, leading to graft-versus-host disease.[158]^25 Elevated levels of serum IFNγ and expansion of peripheral T cells were observed only in the T cell-infused xenograft mice, which was further amplified by TCE treatment ([159]Figures 5C and 5D). In contrast, xenograft mice implanted with NK cells alone showed no significant induction of serum IFNγ nor NK cell expansion in blood ([160]Figures 5C and 5D). However, NKCE treatment activated NK cells to produce IFNγ and increased NK cell infiltration into the tumor microenvironment (TME), without peripheral NK cell expansion ([161]Figures 5C–5E). Lastly, NKCE-IV and NKCE-V induced greater NK cell activation compared to NKCE-III, as shown by the increased frequency of CD69^+ cells ([162]Figure 5F). Overall, these data demonstrated that NKCEs elicit potent antitumor responses and NK cell infiltration in an NSG model. However, due to the technical constraints of the allogenic response associated with T cell infusion alone, a direct comparison with a TCE cannot be made in this model. Figure 5. [163]Figure 5 [164]Open in a new tab huCLL1 NKCEs significantly delay tumor growth in a NOMO1-Luc xenograft model (A) huCLL1 TCE and NKCEs used in the study. (B–F) NSG mice (n = 10) were subcutaneously implanted with a 2:1 mixture of NOMO1-Luc and human Pan T or NK cells and subsequently treated with isotype (1 mg/kg), huCLL1 TCE (0.7 mg/kg), NKCE-III (1 mg/kg), NKCE-IV (0.8 mg/kg), or NKCE-V (1 mg/kg) as indicated by arrows below the x axis. (B) Tumor growth curve. TGI is indicated in the graph. (C) Sera IFNγ at 24 h post-first dose of tool molecules. (D) The concentration of human immune cells (T or NK cells) per microliter (μL) of blood was measured by flow cytometry at day 10 post-tumor implantation. (E) The number of tumor-infiltrating NK cells is shown normalized to tumor volume. (F) Frequency of CD69^+ cells is shown for tumor-infiltrated NK cells. Data are presented as means ± SEM. The data shown are representative of at least two independent experiments. p values were determined by repeated-measures two-way ANOVA followed by Dunnett’s multiple comparisons test (B) or one-way ANOVA followed by Tukey’s multiple comparisons test (E and F) or Mann-Whitney test (C and D). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. We next compare NKCEs and TCEs in in vivo tumor studies using immunocompetent mice. Targeting CD19 on leukemia cells is a clinically validated strategy for the treatment of leukemia.[165]^3 Hence, NKCE and TCE molecules targeting mouse CD19 (muCD19) were generated ([166]Figure 6A). An NKCE (muCD19 NKCE) against muCD19 on a mouse IgG1 (muIgG1) SEFL backbone was developed to deliver mouse WT IL-2 and engage mouse NCR1 with K[D] = 1.4 nM ([167]Figure 6A; [168]Table S3). A HLE muCD19 TCE consisting of an anti-mouse CD19 scFv, and an anti-human CD3 scFv with Fc domain, was generated as a comparative TCE ([169]Figure 6A; [170]Table S3).[171]^26 A mouse IL-2 on a muIgG1 SEFL backbone (muIgG1-IL-2) molecule was also generated to serve as a control to evaluate the effect of IL-2 engagement alone ([172]Figure 6A). Figure 6. [173]Figure 6 [174]Open in a new tab muCD19 NKCEs promote tumor immunity without notable adverse events (A) Design of mouse tool molecules including isotype, muIgG1-IL-2, muCD19 TCE, and muCD19 NKCE. (B–D) huCD3 KI mice (n = 10) with MC38-muCD19 tumors were treated with isotype (1 mg/kg), muIgG-IL-2 (1.15 mg/kg), muCD19 TCE (0.7 mg/kg), or muCD19 NKCE (1 mg/kg) as indicated by arrows below the x axis. (B) Tumor growth curve. (C) The concentration of B220^+ B cells per μL of blood was measured by flow cytometry at day 24 post-tumor implantation. (D) Sera TNF at 4 and 24 h post-first dose of tool molecules. (E–G) Rag2^−/− mice (n = 10) bearing B16F10-muCD19 tumors were treated with isotype (0.5 mg/kg), muIgG-IL-2 low (0.3 mg/kg), muIgG-IL-2 high (0.6 mg/kg), muCD19 NKCE low (0.25 mg/kg), or muCD19 NKCE high (0.5 mg/kg) at days 8 and 13 after tumor implantation, as indicated by arrows below the x axis. (E) Tumor growth curve. (F and G) Sera cytokines at 4 h post-first dose of tool molecules from combination of three experiments, (F) IFNγ and (G) TNF. (H and I) 4 days post-first dose of tool molecules, (H) the number of tumor-infiltrating NK cells are shown, normalized to tumor volume. (I) Expression of Granzyme B on tumor-infiltrated NK cells. Data are presented as means ± SEM. The data shown are representative of at least two independent experiments. p values were determined by repeated-measures two-way ANOVA followed by Dunnett’s multiple comparisons test (B and E) or one-way ANOVA followed by Tukey’s multiple comparisons test (C, D, and F–I). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. To compare efficacy and cytokine release between CD19 engagers, immune-competent human CD3ε knockin (KI) mice were implanted with an MC38 tumor cell line expressing muCD19 ([175]Figure S6A).[176]^26 Tumor-bearing mice treated with molecules muIgG1-IL-2, muCD19 NKCE, and muCD19 TCE lead to approximately 70% TGI compared to the control muIgG1-treated group ([177]Figure 6B). However, despite similar tumor inhibition, analysis of pharmacodynamic markers and immune cell profiling revealed distinct in vivo effects among molecules tested. Consistent with previous results,[178]^26 a high dose of muCD19 TCE resulted in the near-complete depletion of endogenous CD19 expressing B cells in the periphery ([179]Figures 6C and [180]S6A). In contrast, muIgG-IL-2 and muCD19 NKCE treatments did not deplete endogenous B cells, which may be attributed to the significantly lower CD19 expression relative to MC38-muCD19 cells ([181]Figure S6A) or interaction of inhibitory receptors on NK cells and major histocompatibility complex class I molecules on normal CD19^+ B cells.[182]^27 Additionally, muCD19 TCE induced rapid transient induction of cytokines including TNF, IFNγ, IL-6, and growth-regulated alpha protein (KC/GRO) at 4 h, which declined at 24 h ([183]Figures 6D and [184]S6B–S6D). Consistent with in vitro results, significantly lower sera cytokines were detected post-24 h treatment of muCD19 NKCE relative to muIgG1-IL-2 or muCD19 TCE ([185]Figures 6D and [186]S6B–S6D). Furthermore, splenomegaly was apparent in mice treated with muIgG1-IL-2 ([187]Figure S6E). In contrast, reduced spleen size was evident after muCD19 TCE treatment, likely due to depletion of B cells ([188]Figures 6C and [189]S6E). Lastly, muCD19 NKCE demonstrated increased NK cell infiltration within the TME, whereas muIgG1-IL-2 and muCD19 TCE raised the CD8^+ T cell/regulatory T cells (Treg) ratio ([190]Figures S6F and S6G). Thus, NKCEs demonstrated comparable efficacy to IL-2 and TCE treatment in immune-competent mice but with more favorable safety profile. Given the similar antitumor efficacy seen in the MC38-muCD19 mouse model between muCD19 NKCE and muIgG1-IL-2 treatment, and the potential for IL-2 to activate T cells, we sought to use another model to determine whether engagement of NK cells alone is sufficient to drive tumor rejection. To test this, we implanted an immunologically cold B16F10 tumor cell line expressing muCD19 into Rag2^−/− mice, which contain NK cells but lack T and B cells ([191]Figure S6A).[192]^28^,[193]^29 High dose of muCD19 NKCE significantly delayed B16F10-muCD19 tumor growth compared to isotype control and outperformed muIgG1-IL-2 treatments ([194]Figure 6E). In contrast to MC38-muCD19 studies ([195]Figure S6B), muCD19 NKCE induced a low but dose-dependent increase in sera IFNγ and TNF production at 4 h post-injection in B16F10 tumor-bearing Rag2^−/− mice ([196]Figures 6F and 6G). These differences may be attributed to the lack of T cells and higher NK cell numbers seen in Rag2^−/− mice.[197]^30 Furthermore, flow cytometry analysis 4 days after a single dose of muCD19 NKCE or muIgG1-IL-2 treatment revealed a marked increase in tumor-infiltrating NK cells ([198]Figure 6H). While high doses of both treatments had similar effects, low-dose muCD19 NKCE was superior to muIgG1-IL-2 in promoting NK cell infiltration ([199]Figure 6H). Both muCD19 NKCE and muIgG1-IL-2 enhanced Granzyme B expression on NK cells ([200]Figure 6I). In conclusion, tumor-targeted NKCEs can delay tumor growth by boosting the infiltration and function of NK cells, with a reduction in the adverse events seen in IL-2 treatment alone or TCE treatment. Discussion Most efforts in cancer immunotherapy have focused on enhancing the T cell response against tumor cells. However, TCEs and CAR-T therapy have been linked to dose limiting adverse events, such as CRS.[201]^31^,[202]^32 TNF is believed to be the primary cytokine contributing to CRS.[203]^8 Production of TNF by activated T cells stimulates monocytes and macrophages to become the dominant cellular source of IL-6 release during TCE therapy.[204]^8 In turn, IL-6 may further amplify activation of other immune cells (e.g., B cells, T cells, and mast cells)[205]^33 and induce C-reactive protein release via the acute phase liver response[206]^34 and synergy with TNF to activate pericytes to reduce the blood-brain barrier.[207]^35 Therefore, efforts to limit TNF and overall cytokine production are key for effective immunotherapy. NK cells share similar mechanisms for lytic killing of tumor cells with T cells but induce fewer cytokines, such as TNF, upon activation.[208]^7 These properties have led to the exploration of NK cells as an alternative immune cell type to engage and deliver to the TME as an oncology therapy. This is supported by recent NK cell therapy studies that demonstrated lower adverse events than CAR-T therapies in multiple clinical trials.[209]^36 Our studies identify a potential mechanism for the observed reduction in TNF by NK cells relative to activated T cells. We identified inherent cell-intrinsic differences between NK and T cells in TNF trafficking and processing of membrane-bound TNF following chemical activation using stimuli that bypass the need to engage activating receptors. Although primarily regulated at the biosynthesis level, evidence suggests that intracellular trafficking can be rate limiting for the secretion of TNF.[210]^37 Cellular differences in TNF secretion between NK and T cells were also observed during benchmarking studies using NKCE and TCE molecules that activate co-receptors. It is possible that the differences observed with engager molecules may be attributed to intrinsic cell-type differences in activation receptor abundance or differences in signaling thresholds. For instance, CD3 expression on T cells is significantly higher than NCR1 and CD16 expression on NK cells. Both CD3 signaling in T cells and CD16/NCR1 signaling in NK cells require ITAMs for activation and TNF release.[211]^38 However, T cells exhibit greater efficiency in activation due to their larger stoichiometry of 10 activating ITAMs, compared to up to 6 ITAMs within NK cells.[212]^39^,[213]^40^,[214]^41 Additionally, engager affinity for activating receptors may contribute to these differences, as seen in studies that have modified a TCE’s on/off rate for CD3, resulting in effective tumor cell killing while showing reduced cytokine release profiles.[215]^42 Nevertheless, our studies indicate that NK cells and the NKCE molecules used show a significantly reduced TNF release profile both in vitro and in vivo compared to TCEs. By generating various NKCE molecular formats and comparing them against IL-2 and TCEs, we have revealed design strategies to enhance the efficacy of NKCEs. One such strategy is increasing the ability of NKCEs to engage not just NCR1 or CD16 but also both receptors simultaneously, which provides a synergistic effect in increasing NKCE potency. Additionally, benchmarking studies comparing NKCEs to TCEs or soluble IL-2 revealed the need to add IL-2 to NKCE molecules to achieve increased Emax and potent EC50 tumor killing properties, on par with TCEs. Transcriptional and fluorescence-activated cell sorting (FACS) analysis showed a remarkable upregulation of cytotoxic and activating receptors in NK cells treated with NKCEs that engage IL-2R, but not IL-2-unincorporated NKCEs. This was accompanied by no further increase in TNF. Therefore, NKCEs incorporating IL-2 may enhance antitumor efficacy without compromising safety. Indeed, this was seen in our MC38 tumor studies in which non-targeted IL-2 resulted in equivalent TGI as IL-2-incorporated NKCEs but resulted in higher cytokine release, splenomegaly, and a lack of increase in NK cell migration into the TME. The use of incorporating IL-2 into NKCEs is further supported by a recent efficacy and safety study of tetraspecific NKCEs such as antibody-based natural killer cell engager therapeutics (ANKET), which contains an IL-2 variant in non-human primates.[216]^43 Although NKCE-V reported here and ANKET engage the same receptors (NCR1, CD16, and IL-2R) on NK cells and bind NCR1 with similar affinity, NKCE-V encodes a WT IL-2 while ANKET encodes an IL-2 variant with reduced affinity for CD25, which is an IL-2R subunit highly expressed by Tregs and memory T cells.[217]^43 The use of this IL-2 variant is believed to avoid undesired expansion of Tregs within the tumor environment and has the potential to boost antitumoral efficacy.[218]^44 However, our studies and others lack benchmark comparisons between WT and IL-2 variant encoded NCKEs to fully elucidate any potential for therapeutic differentiation.[219]^43 Indeed, both WT IL-2 and CD25-biased IL-2 have recently been shown to exhibit better antitumor immunity than CD25-attenuated IL-2.[220]^45 Furthermore, IL-15, which does not activate CD25^+ Tregs, has also been explored for the development of NKCEs to enhance NK cell activation and antitumor activity.[221]^46 Despite the difference, our study and others support the notion of combining NKCEs with IL-2, offering a differentiated safety profile from non-targeted IL-2 therapies and TCEs.[222]^43^,[223]^47 In summary, our work provides an in-depth analysis of common and different signaling pathways between engaging NK and T cells during antitumor responses. By generating multiple formats and performing comparative studies against TCEs and IL-2 in three separate in vivo models, we demonstrate the efficacy and potential advantages of different NKCE formats. Our study provides a framework for the further improvement of NKCEs and future development of therapeutic candidates. Limitations of the study Study limitations may impact data interpretation. Despite NKCE and TCE molecules being designed to have similar affinities for tumor antigens and respective co-receptors, their different geometric formats result in molecular variances of target affinities. These variances are less than 2-fold for co-receptors among NKCE molecules and less than 4-fold between TCE and NKCE co-receptors. Additionally, differences exist in NK and TCR target densities, which may influence benchmarking studies. This study uses WT IL-2 protein in our NKCE and control tool molecules. While the study indicated the benefit of IL-2 targeted to NK cells compared to non-targeted IL-2 in tumor inhibition in the RAG2^−/− tumor model, a bystander effect of NKCE-derived IL-2 on other immune cells cannot be excluded in immune-competent models such as the MC38 study, which showed equivalent efficacy to non-targeted IL-2. Further studies comparing NKCEs with reduced affinity IL-2 variants or biased NK agonism are necessary to address this question. We found no evidence of tumor cells directly influencing NK cell survival or secretion of NK survival factors (i.e., IL-2) by either NOMO1-Luc or JIMT1-Luc, but we did not systematically rule out the effect of other tumor lines used in this study on influencing NK cell viability during co-cultures conditions. Although IL-2 has a short half-life with low physiological levels,[224]^48 our in vivo studies did not evaluate the contribution of endogenous IL-2 on NKCE efficacy. Additional in vivo benchmarking studies comparing NKCEs with or without incorporated IL-2 are necessary to address this matter. Lastly, our study found that NK cells, compared to T cells, show less trafficking of TNF to the cell membrane and release of membrane-bound TNF. Further studies are needed to understand this mechanism. Resource availability Lead contact All inquiries for further information regarding this study should be directed to and will be fulfilled by the lead contact, Weiwen Deng (wdeng01@amgen.com). Materials availability Materials will be made available to the scientific community by completion of a material transfer agreement. Requests can be made at [225]www.amgen.com/partners/academic-collaborations/new-requests. Data and code availability * • The raw sequence data reported in this paper, along with a processed FPKM count table, are available through GEO: [226]GSE264122 . * • This paper does not generate original code. All data associated with this study are presented in the paper or [227]supplemental information. * • Any additional information required to reanalyze the data reported in this paper is available from the [228]lead contact upon request. Acknowledgments