Comparative Evaluation of Synthetic Cytokines for Enhancing Production and Performance of NK92 Cell-Based Therapies

Abstract

Off-the shelf immune cell therapies are potentially curative and may offer cost and manufacturing advantages over autologous products, but further development is needed. The NK92 cell line has a natural killer-like phenotype, has efficacy in cancer clinical trials, and is safe after irradiation. However, NK92 cells lose activity post-injection, limiting efficacy. This may be addressed by engineering NK92 cells to express stimulatory factors, and comparative analysis is needed. Thus, we systematically explored the expression of synthetic cytokines for enhancing NK92 cell production and performance. All synthetic cytokines evaluated (membrane-bound IL2 and IL15, and engineered versions of Neoleukin-2/15, IL15, IL12, and decoy resistant IL18) enhanced NK92 cell cytotoxicity. Engineered cells were preferentially expanded by expressing membrane-bound but not soluble synthetic cytokines, without compromising the radiosensitivity required for safety. Some membrane-bound cytokines conferred cell-contact independent paracrine activity, partly attributable to extracellular vesicles. Finally, we characterized interactions within consortia of differently engineered NK92 cells.

Introduction

Cell-based cancer immunotherapies have been transformative for certain patients, establishing the importance of addressing challenges that limit their widespread use. Cell-based cancer immunotherapies comprise immune cells with enhanced capacity to kill tumor cells.¹ Currently, chimeric antigen receptor (CAR) T cells are the most developed and well-studied cell-based cancer therapies.² CAR-T cells can achieve long-term remissions in patients who are unresponsive to multiple lines of chemotherapy, marking a major advance in cancer therapy.²

Widespread use of CAR-T cells is limited by high cost and complexity of production.³ CAR-T cells must be genetically identical to the patient to avoid graft-versus-host disease, a complication where grafted T cells attack healthy tissue.⁴ Therefore, CAR-T cells are typically produced in an autologous fashion—using the patient's T cells—which requires between 9 and 14 days and involves complex supply chains.^5,6 While mitigating these challenges is the subject of ongoing development and research, it also highlights opportunities for developing complementary approaches.

Natural killer (NK) cell-based therapies are promising alternatives to T cell-based therapies. They confer many comparable benefits (e.g., ability to kill tumor cells) but, unlike T cells, do not need to be genetically identical to the patient, and therefore have the potential to be an “off the shelf” therapy, derived from either allogeneic donors or immortalized cells.⁷ Like CAR-T cells, CAR-NK cells lyse tumor cells and confer durable remissions in patients.⁸

The NK cell-based therapies have the additional benefits of causing only mild side effects and naturally recognizing markers of tumor cells, such as low expression of major histocompatibility complexes.^8,9 There are multiple sources for NK cells, including the immortalized NK92 cell line, which is attractive for some applications. NK92 cells, which are derived from a non-Hodgkin lymphoma, can replicate for long periods of time, demonstrate NK cell-like behavior, express high levels of cytotoxicity activating receptors, and are relatively easy to engineer compared to primary NK cells.^10,11

After irradiation to prevent tumorigenesis, NK92 cells are safe to administer and show therapeutic promise in clinical studies.¹² Notably, NK92 cell therapies are substantially cheaper to produce than CAR-T cells (less than $20,000 vs. an estimated $58,2000 cost of goods per treatment).^10,13 Finally, NK92 cells can be repeatedly genetically modified, enabling bioengineers to deliver transgenic cargo above the 10 kb limit conferred by standard (e.g., lentiviral) vectors.¹⁴ The increased capacity for modification allows for safety improvements, efficacy improvements, and implementation of sophisticated genetic programs for conferring customizable cellular functions.

Synthetic cytokine (SC) expression may improve production and efficacy of NK92 cell-based therapies. NK92 cell expansion and anti-tumor efficacy depend on signaling from cytokines, proteins that coordinate immune responses.^11,15 The SCs are analogs of natural cytokines engineered with desirable properties, including improved immune-activating potency and removal of undesirable paradoxical immunosuppressive effects.¹⁶

The SC expression may facilitate NK92 cell-based therapy biomanufacturing by enabling selection and maintenance of engineered cells in recombinant cytokine free culture, simplifying biomanufacturing protocols and requirements.¹⁷ The SC expression could potentially lead to higher NK92 cell efficacy after administration, enable NK92 cells to activate surrounding endogenous immune cells, and eliminate the need for systemic co-administration of cytokines, which are only effective at high doses that can lead to toxic side effects.^18,19 To date, some individual SCs have been shown to be effective in enhancing NK92 cell production and/or performance,^20–23 but a systematic, comparative analysis across SCs is needed to guide future bioengineering—addressing this gap motivates this investigation.

In this study, we evaluated and compared the impact of expressing six SCs on properties relevant to NK92 cell-based therapy manufacturing (selection and expansion of engineered cells in cytokine-free media), cytotoxicity, and paracrine activity.^21,24–28 We evaluated the impact of SC expression on NK92 cell expansion and cytotoxicity in hypoxia, a feature of solid tumors that limits the efficacy of NK cell-based therapies.^29,30

Membrane bound, but not soluble, SCs enable selection of engineered cells in cytokine-free media, and effects on growth and cytotoxicity varied as a function of cytokine choice and environmental condition. Some SCs enabled paracrine activation of NK92 cell growth or cytotoxicity, and mixing NK92 cell lines engineered in different ways yielded useful ensemble behaviors. The SCs have differential properties and relative advantages that can be used to inform the choice of SC in future development in NK92 cell-based therapies.

Results

Expression and bioactivity of synthetic cytokines

A panel of soluble synthetic cytokines (sSCs) was selected and epitope tagged, then validated for expression and bioactivity on NK92 cells (Fig. 1). The first sSC—Neoleukin-2/15 (Neo2/15)—is an analog of IL2 that preferentially activates pro-inflammatory cells and, when produced by NK92 cells, stimulates co-cultured immune cells.^22,25 The second sSC—engineered IL15 (eIL15)—is a fusion of IL15 and the IL15Rα sushi domain that acts as a super agonist on the IL15 receptor,²⁷ activates NK92 cell cytotoxicity,³¹ and is being evaluated in combination with an immune checkpoint inhibitor in a phase II clinical trial (NCT05256381).

FIG. 1.

Modified SC constructs are expressed and bioactive.

(A) Evaluation of Myc-tagged Neo2/15, 3xFLAG-tagged Neo2/15, eIL-12, eIL-15, and 3xFLAG-tagged DR18 expression in HEK293FT cells.

(B) Evaluation of bioactivity of SCs (Neo2/15, eIL12, eIL15) in HEK293FT cell-conditioned media. In each case, HEK293FT cells were transiently transfected with an empty vector or a vector encoding the indicated SC. Fresh cell medium was conditioned by transfected HEK293FT cells for 24 h, starting 48 h after transfection. Conditioned supernatant was harvested and clarified by centrifugation and filtering to remove residual cells and debris. The 4 × 10⁴ NK92 cells were then cultured in 50% conditioned medium, then counted after 3 days to assess cell growth. Positive controls include NK92 cells cultured in 50% empty vector conditioned medium supplemented with 100 IU/mL rIL2, 10 ng/mL rIL12, or 10 ng/mL rIL15 as indicated. ****p < 0.0001, **p < 0.01, *p < 0.05 (one-way ANOVA, Tukey's HSD). n = 4 (Neo2/15), n = 3 (eIL12), n = 3 (eIL15).

(C) Evaluation of bioactivity of DR18 in transduced NK92 cells. NK92 cells were transduced with an empty BB lentiviral expression vector encoding no transgene or that same vector driving expression of 3xFLAG-DR18. NK92 cell lines were cocultured with K562 cells for 20 h, with and without 5 IU/mL IL2. Live K562 cell counts were measured by flow cytometry and used to calculate %K562 growth inhibition, which is defined by Equation (1) (Materials and Methods section). Points indicate mean values, and error bars indicate standard deviation. n = 3. AUC analysis of each condition was performed. Error bars represent SEM and ****p < 0.0001 (two-way ANOVA, Tukey's HSD). See Supplementary Note 1 for full ANOVA results. ANOVA, analysis of variance; AU, arbitrary units; AUC, area under the curve; BB, backbone; DR18, decoy-resistant interleukin18; eIL, engineered interleukin; HSD, honest significant difference; Neo, neoleukin; SC, synthetic cytokine; SEM, standard error of the mean; rIL, recombinant interleukin.

The third sSC—engineered IL12 (eIL12)—is a fusion of the IL12 p35 and p40 monomers (preventing their association into potentially inhibitory cytokines)³² that potentiates anti-tumor responses in vivo,²⁰ and, when expressed by NK92 cells, stimulates co-cultured immune cells.²⁰ The fourth sSC—decoy-resistant IL18 (DR18)—is a mutated variant of IL18 that does not bind to an inhibitory protein, enhancing its ability to stimulate anti-tumor responses in vivo and activate NK cells.²⁶ The sSC transgenes were codon-optimized for expression in human cells with features chosen to enhance stability and add or remove epitope tags as needed (Materials and Methods section).

To validate the expression of modified sSC constructs by human cells, HEK293FT cells were transiently transfected with sSC constructs and analyzed by western blot (Fig. 1A). Neo2/15 was only expressed when tagged on its N′ terminus, indicating that Neo2/15 is destabilized when its C terminus is altered. Neo2/15 was also stably expressed when the N terminal Myc tag was changed to a 3xFLAG tag. A 3xFLAG-tagged Neo2/15 (hereafter referred to as Neo2/15) was used for all subsequence experiments. The eIL12, eIL15, and N′ terminus-tagged DR18 constructs were all expressed by HEK293FT cells. Viable constructs were carried forward for further validation.

After validating that modified sSCs were expressed, they were tested for bioactivity on NK92 cells. The native analogs of Neo2/15 (IL2 and IL15), eIL15 (IL15), and eIL12 (IL12) are sufficient to induce NK92 proliferation in the absence of other cytokines,^11,33,34 so it was hypothesized that these sSCs, if bioactive, would also be sufficient to induce NK92 cell proliferation. To test this property, sSC-conditioned medium was evaluated for its ability to induce NK92 cell growth in the absence of recombinant IL2, which is usually required for NK92 cell proliferation.

All three sSCs induced growth, indicating bioactivity (Fig. 1B). As the native analog of DR18 (IL18) enhances NK92 cell cytotoxicity in the presence of small amounts of IL2,³⁵ DR18 should do the same if bioactive. NK92 cells transduced to express 3xFLAG-tagged DR18 (hereafter referred to as DR18) were tested for their ability to reduce K562 B cell lymphoma cell counts in co-culture; throughout this study, we use a rigorous metric for quantifying cytotoxicity by comparing the number of live K562 remaining after co-incubation with NK92 cells to the number of live K562 that would be present if no NK92 cells were added [Eq. (1), Materials and Methods section].

DR18 expression significantly enhanced NK92 cell-mediated K562 growth inhibition, particularly in the presence of 5 IU/mL IL2, indicating that DR18 induces autocrine and paracrine bioactivity on NK92 cells (Fig. 1C). Altogether, these results indicate that our constructs were functional and fit for subsequent evaluation.

Selective effects conferred by synthetic cytokine expression

We next investigated whether SC expression enables selection of genetically engineered NK92 cells. We evaluated the previously validated sSCs, as well as two membrane-bound synthetic cytokines (mbSCs). The first mbSC—membrane-bound IL2 (mbIL2)—comprises a fusion of IL2 to IL2Rβ and enhances NK92 cell cytotoxicity and survival.²¹

The second mbSC—membrane-bound IL15 (mbIL15)—is a fusion of IL15 to IL15Rα, enhances NK92 cell proliferation in vitro, and is related to a membrane-bound IL15 construct that enhanced the proliferation and in vivo cytotoxicity of NK92 cells.^23,28,36 Genetic engineering techniques often insert transgenes into a limited fraction of cells, especially as the transgenic cargo increases in size,³⁷ which can necessitate positive selection (enrichment of genetically engineered cells) to develop an engineered cell product.

In contrast, negative selection (depletion of genetically engineered cells or selective expansion of cells that have lack the transduced genes) is the undesired outcome that would pose challenges to developing an engineered cell product. As IL2 and IL15 confer survival of NK92 cells in the absence of IL2,³⁴ while IL18 and IL12, despite activating NK92 cell cytotoxicity,^35,38 also induce NK92 cell apoptosis when present at higher concentrations,^39,40 we hypothesized that only synthetic counterparts of IL2 or IL15 would impart positive selective pressures on NK92 cells, while DR18 and eIL12 would exert negative selective pressure.

We first evaluated whether the expression of SCs would exert selective pressure when starting from a population with a low percentage of transduced cells. NK92 cells were transduced with lentiviral vectors encoding SCs, cultured in cytokine-free media for 8 days, and then evaluated for changes in the percentage of engineered cells (Fig. 2A). As Neo2/15 and DR18 lack known signal peptides, which usually direct cytokine trafficking to cellular compartments in which they can bind cytokine receptors,⁴¹ we evaluated Neo2/15 and DR18 constructs with or without an additional N′ terminal IgE signal peptide (notated as sp^IgE vs. sp^null, respectively).

FIG. 2.

Expression of membrane-bound but not soluble SCs confers positive selection on NK92 cells.

(A) Schematic describing NK92 cell engineering and subsequent selection.

(B, C) Evaluating the potential of SC expression to confer positive selection for mbSCs (B) and sSCs (C). NK92 cells transduced to express SCs were cultured with or without 100 IU/mL rIL2 for 8 days. Transduced cells were identified by the expression of relevant markers (miRFP720 for mbSCs, mNeonGreen for sSCs).

(D) Transgene expression frequency distribution for NK92 cells expressing mbSC after 6 weeks of culture without IL2.

(E) Quantification of transgene expression by mbSC transduced NK92 cells in (D). Cells positive for miRFP720^pos were gated on, and magnitude of miRFP720 signal was analyzed. Bar height indicates mean signal. Error bars represent standard deviation. ****p < 0.0001 (two-tailed unpaired t-test). Cells were analyzed from one sample, each.

(F, G) Evaluating negative selection. Transgene expression was monitored for 52 days. Since cells were sorted on different days but analyzed together, each time point represents a range of days post sorting (see Supplementary Note 2 for details). As described in the text, eIL12-expressing lines recovered slowly and were not analyzed at day 0. (F) is an analysis of the histograms in (G). mbSC, membrane-bound SC; sSC, soluble SC.

Since, in preliminary work (not shown), we observed that eIL12 expression downstream of the EF1α promoter induced death of transduced NK92 cells, we evaluated several levels of eIL12 expression using various upstream open reading frames,⁴² which led to the development of one line capable of sustained expression of eIL12 when cultured with rIL2 (Supplementary Fig. S1). After 8 days of culture in cytokine-free media (i.e., without supplementation with recombinant cytokines), only mbIL2 and mbIL15 expression conferred strong positive selection of transduced cells (as compared with populations cultured with rIL2) (Fig. 2B).

Expression of soluble cytokines did not serve as effective selectable markers, although expression of sp^IgENeo2/15 and eIL15 led to slight enrichment of transduced cells (Fig. 2C). After 6 weeks of culture without rIL2, the mbIL15 selection conferred greater purity (i.e., frequency of transduced cells in the final population) and higher transgene expression compared with mbIL2 (Fig. 2D, E). This outcome may reflect different magnitudes of survival signaling conferred by these mbSCs. A lack of selective advantage was expected for DR18 and eIL12, as neither cytokine is expected to enable long-term NK92 cell survival.

A lack of selective advantage for Neo2/15 and eIL15 may be explained by paracrine expansion of non-transduced cells, potentially including competition between transduced and non-transduced cells for a common pool of secreted factors.

We next evaluated whether the expression of sSCs exerted negative selective pressure when starting from a population enriched for transduced cells, as manifested by silencing of the sSC transgenes and expansion of this undesirable population. NK92 cells engineered to express these factors were enriched by fluorescence activated cell sorting (FACS), and subsequent changes in the proportion of engineered cells were tracked over multiple months. eIL15 did not impart negative selective pressures on engineered NK92 cells but Neo2/15 and DR18 did, particularly when cytokine secretion was enhanced with the IgE signal peptide (Fig. 2F, G).

Possible explanations include that the latter sSCs may be toxic to NK92 cells when present above a certain threshold, they may preferentially enhance the expansion of co-cultured cells, or a combination of these effects may exist. We could not similarly evaluate eIL12-expressing NK92 cells at the earliest time point (“Day 0”; 10 days after FACS enrichment) because cells were present at low number and were slower to recover than other lines, although a stable, low transgene expression frequency was eventually reached (∼40%).

This pattern may reflect the toxicity of the eIL12 transgene and its effects after stressors such as FACS. In aggregate, these observations indicate that mbIL2 and mbIL15 can confer positive selection by enabling preferential expansion of engineered NK92 cells, whereas eIL15 neither substantially enhances nor diminishes the relative proportion of engineering NK92 cells, and that Neo2/15, eIL12, and DR18 may actually diminish the proportion of engineered NK92 cells when expressed above a certain level.

Effects of membrane-bound synthetic cytokines on NK92 cells challenged with hypoxia and irradiation

In practice, NK92 cell-based therapies will face challenges that could modulate the effects conferred by synthetic cytokines. In particular, NK92 cells must be irradiated before administration to prevent tumorigenesis,¹² and the hypoxic tumor microenvironment may modulate NK92 cell cytotoxicity and survival.³⁰ We, thus, investigated how signaling from membrane-bound synthetic cytokine (mbSC) expression impacted NK92 cell growth and when treated with 10 Gy of irradiation (a clinically relevant dose),¹² or under hypoxia.

We first examined the effects of mbSC expression on NK92 cell expansion in the absence of exogenous cytokine. Both mbIL2 and mbIL15 induced the formation of large NK92 cell clumps, a sign of cell health and proliferation. The mbIL15 NK92 cells began showing signs of overgrowth (clump spreading and degradation) sooner than did mbIL2 NK92 cells, indicating that mbIL15 NK92 cells either grew more quickly or were more sensitive to overgrowth (Fig. 3A and Supplementary Videos S1–Fig. 3A and S3).

FIG. 3.

mbSC expression enhances the growth and cytotoxicity of NK92 cells under challenges.

(A) Illustrative images of NK92 cell growth effects. NK92 cells transduced to express mbSCs were cultured in cytokine-free media and imaged via time lapse microscopy (see also Supplementary Videos S1–S3).

(B) Effects of mbSC expression on NK92 cell growth in normoxia (∼21% O₂), hypoxia (1% O₂), or after irradiation (10 Gy). Sample labels in this panel also apply to (D).

(C) Effects of mbSC expression on NK92 cell cytotoxicity against K562 tumor cells. NK92 cells were cultured without IL2 for 24 h, then mixed with fluorescent K562 cells at a ratio of 3.125 NK92:1 K562 for 20 h. K562 cells (mScarlet-I^pos) were counted using image-based cell-cytometry.

(D) Effects of mbSC expression on NK92 cell cytotoxicity as a function of effector to target (E:T) ratio and oxygen tension. NK92 cells were cultured without IL2 for 24 h, then mixed with fluorescent K562 cells at varying effector (NK92) to target (K562) ratios (E:T ratios) and incubated at oxygen tensions indicated. After 20 h, live K562 cell counts were obtained by flow cytometry. AUC analysis of K562 inhibition curves was performed. *p < 0.05, **p 0.01, ***p < 0.001 (one-way ANOVA, Tukey's HSD). See Supplementary Note 1 for full ANOVA results.

(E) Effects of mbSC expression on NK92 cell cytotoxicity following irradiation. NK92 cells were cultured without IL2 for 24 h and irradiated as indicated, then mixed at varying E:T ratios with K562 cells, and incubated in ambient oxygen for 20 h. Live K562 cell counts were measured by flow cytometry. For all line graphs, points represent means and error bars represent standard deviation, n = 3.

Expression of either mbSC increased the growth of NK92 cells in normoxia (21% O₂), with mbIL15 trending marginally higher than mbIL2, and marginally improved persistence in hypoxia (1% O₂) (Fig. 3B). Neither mbSC enhanced cell persistence after irradiation with 10 Gy (Fig. 3B), indicating that mbIL2 and mbIL15 are not sufficient to confer NK92 cell resistance to irradiation. This contrasts with prior findings that mbIL2 slightly enhanced cell persistence after irradiation,²¹ which may reflect the higher rate of irradiation used in the current study (208 cGy/min vs. 83 cGy/min used in prior work), a factor known to alter cell apoptosis in response to irradiation.⁴³

We next examined whether mbSC expression altered NK92 cell cytotoxicity. For both mbIL2- and mbIL15-expressing NK92 cells, we directly observed cytolysis of K562 B cell lymphoma cells (Supplementary Videos S4, S6 and S7). These cell lines exhibited similar kinetics of activity against K562 over 20 h compared with each other, with both being more effective than control NK92 cells (Fig. 3C). Across multiple effector to target (E:T) ratios, when compared with control NK92 cells, both mbSCs significantly enhanced NK92 cell-mediated K562 growth inhibition in normoxia, while in hypoxia, only mbIL15 significantly enhanced K562 growth inhibition (Fig. 3D).

A possible contributing factor is hypoxia-induced acidification of the media, which reduces IL2 signaling.⁴⁴ Expression of either mbSC evaluated here enhanced NK92 cell cytotoxicity after cells were irradiated, improving potential clinical utility (Fig. 3E). Notably, mbIL2 NK92 cells showed greater variations in potency against K562 across repeated experiments, which may indicate variations in cell state that are not attributable to known biology and warrant mechanistic investigations in future studies.

Altogether, these results show that mbSC expression is a viable strategy for enhancing NK92 cell growth and cytotoxicity in normoxia and is compatible with irradiation-based protocols to halt NK92 cell expansion while retaining cytotoxicity.

Effects of soluble synthetic cytokines on NK92 cells challenged with hypoxia and irradiation

We next investigated how soluble synthetic cytokine signaling influenced NK92 cell expansion and cytotoxicity against K562 cells in various conditions of interest, focusing our investigation on the four soluble synthetic cytokines (sSCs) whose bioactivity was validated in Figure 1 (Neo2/15, eIL12, eIL15, and DR18). As recombinant protein monotherapies, these sSCs (or similar constructs) can stimulate NK92 cell cytotoxicity in vitro (eIL15),³¹ and/or activate anti-tumor endogenous immune cell responses in vivo (Neo2/15, eIL12, eIL15, DR18).^25,26,45,46

When produced by NK92 cells, select sSCs can activate proliferation and/or cytotoxicity of co-cultured non-NK92 cells (eIL12, Neo2/15).^20,22 However, the effects of engineering NK92 cells to express sSCs on their own proliferation and cytotoxicity are not well defined. To address this gap in knowledge, we systematically tested how sSC expression affected NK92 cell expansion and cytotoxicity, including evaluations in hypoxia and after irradiation.

Based on the effects of their natural analogs and prior literature, we expected that, in standard conditions, Neo2/15 and eIL15 would enhance growth and cytotoxicity,^22,34 while eIL12 and DR18 would only enhance cytotoxicity.^11,35,38 We further expected hypoxia and irradiation to limit the effects of sSC expression on growth and cytotoxicity.^47,48

We first tested whether sSC expression induced NK92 cell expansion. When cultured without IL2 in normoxia, Neo2/15 expression led to the highest cell counts after incubation in normoxia; eIL15 also supported NK92 cell expansion to a lesser degree, which might reflect different affinities of the two cytokines for their receptor (Fig. 4). No sSC improved NK92 cell persistence or growth in hypoxia by day 4, though some permitted mildly improved survival on day 2 (Fig. 4). All sSC-expressing cell lines remained sensitive to irradiation (Fig. 4), although sp^IgENeo2/15 conferred marginally higher survival (Fig. 4A).

FIG. 4.

sSC expression modulates the growth of NK92 cells under challenges.

(A–D) Effects of sSC expression on NK92 cell growth in ambient oxygen (∼21% O₂) or hypoxia (1% O₂) or after irradiation (10 Gy) (and then grown in ambient oxygen). For all line graphs, points represent means and error bars represent standard deviation, n = 3. To aid in comparison, growth of sSC BB NK92 cells was plotted on (A–D), but represent the same set of experiments.

We next investigated how sSC expression modulates NK92 cell cytotoxicity. NK92 cell-mediated cytolysis was visualized using time-lapse microscopy for each sSC-expressing NK92 cell line to validate that engineered NK92 cells remained cytotoxic toward K562 cells (Supplementary Videos S4, S8–S14). Across multiple effector to target (E:T) ratios, sSC-expression increased NK92 cell-mediated K562 growth inhibition in both normoxia and hypoxia (Fig. 5A–E).

FIG. 5.

sSC expression modulates the cytotoxicity of NK92 cells under challenges.

(A–E) Effects of sSC expression on NK92 cell cytotoxicity as a function of effector to target (E:T) ratio and oxygen tension. NK92 cells were cultured without IL2 for 24 h, then mixed with fluorescent K562 cells at varying effector (NK92) to target (K562) ratios (E:T ratios) and incubated at oxygen tensions indicated. After 20 h, live K562 cell counts were obtained by flow cytometry.

(E) Shows analyses of corresponding E:T data series. *p < 0.05, **p < 0.01 (one-way ANOVA, Tukey's HSD). See Supplementary Note 1 for full ANOVA results.

(F) Effects of sSC expression on rate of NK92 cell cytotoxicity. NK92 cells were cultured without IL2 for 24 h, then mixed with fluorescent K562 cells at a ratio of 3.125 NK92:1 K562 for 20 h. K562 cells (mScarlet-I^pos) were counted using image-based cell-cytometry.

(G) Effects of sSC expression on NK92 cell cytotoxicity following irradiation. NK92 cells were cultured without IL2 for 24 h and irradiated as indicated, then mixed at varying E:T ratios with K562 cells, and incubated in ambient oxygen for 20 h. Live K562 cell counts were measured by flow cytometry. For all line graphs, points represent means and error bars represent standard deviation, n = 3.

There was some variation in the rate of K562 cell killing in normoxia, with eIL12, sp^null, and sp^IgENeo2/15 expression leading to the most rapid and prolonged cytotoxicity (Fig. 5F). Irradiation with 10 Gy did not diminish the effects of sSC expression on NK92 cell-mediated K562 growth inhibition (Fig. 5G). Altogether, these observations indicate that, as expected, sSCs analogous to IL2 or IL15 enhance NK92 cell growth, unless cells are challenged with irradiation or hypoxia, and that, excitingly, all sSCs enhance NK92 cell cytotoxicity, even when cells are challenged with irradiation or hypoxia.

Paracrine effects of synthetic cytokine expression on NK92 cell growth and cytotoxicity

We next investigated whether synthetic cytokine (SC) expression confers paracrine signaling that is sufficient to modulate NK92 cell growth or cytotoxicity. Paracrine signaling plays an important role in the efficacy of cell-based therapies.⁴⁹ Paracrine activity can be cell contact-dependent (e.g., cytokine-receptor trans-presentation)^50,51 or contact-independent (e.g., secretion and diffusion of soluble cytokines).

Based on the literature and data in Figures 4 and 5, we expected soluble SCs to confer appreciable contact-independent paracrine effects on NK92 cell growth and/or cytotoxicity that for Neo2/15 and DR18 would be enhanced by the inclusion of IgE signal peptides. As mbIL2 and mbIL15 are membrane tethered and non-transduced cells survived despite cytokine withdrawal when co-cultured with mbIL2 NK92 cells, but not mbIL15 NK92 cells (Fig. 2D), we also expected mbIL2, but not mbIL15, to have contact-dependent paracrine effects.

We first investigated paracrine effects on NK92 cell growth. Because the expression of mbIL2, mbIL15, Neo2/15 (with and without IgE signal peptide), and eIL15 promoted growth in NK92 producer cells, the paracrine activity of these factors was characterized by evaluating the ability to promote the growth of co-cultured NK92 cells expressing no synthetic cytokines (SC^neg NK92 cells). All SC-expressing NK92 cells, including mbIL15 NK92 cells, conferred appreciable paracrine effects on SC^neg NK92 cell growth (Fig. 6A).

FIG. 6.

Select SCs are sufficient to exert paracrine effects on NK92 cell growth or cytotoxicity.

(A) Paracrine effects of SCs on SC^neg NK92 cell growth with cell contact. SC NK92 cells were co-cultured for 3 days with SC^neg NK92 cells, after which SC^neg NK92 cells were counted by flow cytometry. SC^neg NK92 cell fold change was calculated by dividing SC^neg NK92 cell count by number of SC^neg NK92 cells seeded.

(B) Paracrine effects of SCs on SC^neg NK92 cell growth without cell contact. Media were conditioned for 4 days by SC NK92 cells, then clarified to remove cells. SC^neg NK92 cells were cultured in 50% conditioned media for 3 days, then counted by flow cytometry. rIL2 indicates 100 IU/mL rIL2. Control cell data points are duplicated across panels.

(C) Paracrine effects of SCs on SC^neg NK92 cell cytotoxicity without cell contact. Media was conditioned by eIL12 or 3xFLAG-DR18 NK92 cells for 1 day, then clarified to remove cells. As 5 IU/mL rIL2 enhanced effects of DR18 on NK92 cell cytotoxicity (Fig. 1C), 5 IU/mL rIL2 was added to all conditioned media conditions in the 3xFLAG-DR18 paracrine effect assay, only. SC^neg NK92 cells were incubated with fluorescent K562 cells (mScarlet-I^pos) for 20 h in 50% conditioned medium, after which K562 cells were counted by flow cytometry. K562 growth inhibition curves were generated and analyzed. In line graphs, points represent mean and error bars represent standard deviation. n = 3. Bar height indicates mean AUC, error bars indicate SEM. *p < 0.05, **p < 0.01, and ****p < 0.0001 as calculated by a two-tailed t-test (C, eIL12 analysis only), or an one-way ANOVA followed by Tukey's HSD (all other panels). See Supplementary Note 1 for full ANOVA results.

Given this surprising observation with mbIL15, we speculate that during cytokine-withdrawal selection of mbIL15 NK92 cells (Fig. 2D), it is possible that non-transduced cells were not detected because the autocrine mbIL15 signaling potency is stronger than its paracrine activity and this enabled transduced cells to outcompete non-transduced cells, or that the magnitude of paracrine signaling was lower in these prior experiments.

To test whether the observed paracrine signaling is contact-dependent, cell media conditioned by SC-expressing NK92 cells was tested for the ability to induce SC^neg NK92 cell expansion. Interestingly, despite being membrane-tethered, the expression of both mbIL2 and mbIL15 conferred contact-independent effects (Fig. 6B). We speculated that this effect could be mediated by cleavage-mediated release of mbSC components, by mbSC-induced secretion of other factors that can support NK92 cell growth in trans, and/or by extracellular vesicle (EV)-mediated transfer of mbSCs, as EVs are able to mediate signaling between immune cells (these possibilities are further investigated in subsequent experiments).⁵²

Conditioned medium from sp^IgENeo2/15 and eIL15 NK92 cells, but not sp^nullNeo2/15, also significantly supported SC^neg NK92 cell expansion (Fig. 6B). Sp^nullNeo2/15 might be secreted via an unconventional pathway at a level insufficient to support SC^neg cell growth in the absence of cytokine-producing cells,⁵³ although it remains possible that the paracrine effects of sp^nullNeo2/15 are contact-dependent.

We next investigated paracrine effects on NK92 cell cytotoxicity. Because the expression of DR18 and eIL12 signaling enhances cytotoxicity of NK92 cells expressing these cytokines, we investigated paracrine activity of these factors vis-à-vis enhancement of SC^neg NK92 cell cytotoxicity against K562 cells. Since we could not deconvolute K562 killing by SC and SC^neg NK92 cells in a mixed co-culture format, which would be necessary for assessing contact-dependent paracrine effects, only contact-independent paracrine effects were evaluated. Conditioned medium from eIL12, but not DR18 (regardless of signal peptide), NK92 cells significantly enhanced SC^neg NK92 cell-mediated K562 growth inhibition (Fig. 6C).

It is possible that DR18 signaling is sufficient for priming NK92 cell cytotoxicity over long exposures but is insufficient for rapid paracrine activation of NK92 cell cytotoxicity in the absence of other cytokines. DR18 might also act by prolonging the pro-cytotoxic effects of recombinant IL2 after it is withdrawn. Taken together, these observations indicate that paracrine effects of SC expression in NK92 cells are substantial, varied, and sometimes surprising given prior knowledge, motivating subsequent mechanistic investigations.

Mechanistic evaluation of paracrine effects mediated by membrane-bound synthetic cytokines

To elucidate how the expression of membrane-bound synthetic cytokines (mbSCs) conferred contact-independent paracrine effects, we next investigated whether these effects could be mediated by EVs, non-EV soluble factors, or both (Fig. 7). The EVs are a heterogeneous group of membrane-bound vesicles produced by virtually all cell types.⁵⁴ The EVs present some membrane proteins on their surface and can encapsulate cytosolic proteins; the composition of EVs remains an active area of investigation.⁵⁵

FIG. 7.

EVs partly mediate mbIL15 paracrine activity and non-EV components mediate paracrine activity of both mbIL2 and mbIL15.

(A) Schematic of possible mechanisms by which mbSC expression mediates paracrine effects on NK92 cell growth.

(B) Schematic of EV isolation from mbSC NK92 cells.

(C) Schematic of assay used to test how size-based depletion of soluble factors from medium conditioned by mbSC-expressing HEK293FT cells affects the medium's paracrine effects on NK92 cell growth.

(D) EV-mediated paracrine support of NK92 cell growth. Serum-free medium was conditioned by mbSC-expressing NK92 cells for 2 days. The EVs were isolated from conditioned media and added to SC^neg NK92 cells (3.7 × 10⁹ EVs per well). After 3 days, live SC^neg NK92 cells were counted by flow cytometry.

(E) Evaluation of sub-EV contributors to paracrine effects attributable to mbSC expression. HEK293FT cells were transfected with the mbSCs, and media were conditioned for 24 h before depleting EVs and other species using serial filtration through 100 kDa, 50 kDa, and finally 10 kDa molecular weight cutoff filters. The flow through from each filtration step was evaluated for the ability to support SC^neg NK92 cell growth; SC^neg NK92 cells were cultured in 50% filtered conditioned media for 3 days, then quantified by flow cytometry. Error bars indicate standard deviation. For all experiments, ****p < 0.0001 [one-way (D) or two-way (E) ANOVA followed by Tukey's HSD], n = 3. See Supplementary Note 1 for full ANOVA results. EV, extracellular vesicle.

The importance of EVs in intercellular communication, including communication between immune cells, is increasingly recognized.⁵² To explain the surprising aforementioned observations (Fig. 6B), we hypothesized that mbSC signaling could induce NK92 cells to secrete soluble factors that support growth in trans, mbSCs could be cleaved at the membrane to release soluble cytokines, and/or mbSCs could be packaged onto EVs and secreted to subsequently stimulate recipient NK92 cells (Fig. 7A).

We first evaluated whether EVs from mbSC-expressing NK92 cells might contribute to the observed paracrine effects. The EVs were isolated from serum-free medium conditioned by mbSC NK92 cells using ultracentrifugation, followed by size-exclusion chromatography (Fig. 7B). Isolated EVs exhibited appropriate sizes, protein markers, and morphology expected for EVs according to best practice guidelines as published in the EV literature,⁵⁶ validating the EV isolation methodology (Supplementary Fig. S2B–D). The EVs from mbIL15 NK92 cells, but not mbIL2 NK92 cells, were sufficient to support SC^neg NK92 cell growth (Fig. 7D). This could reflect differential trafficking of the mbSC transmembrane domains (IL15Rα vs. IL2Rβ, respectively) to NK92 cell EVs, or differential potency of these factors in this format.

We next investigated whether soluble factors smaller than EVs mediated mbSC paracrine effects. To test this possibility, conditioned media was passed through filters with progressively smaller molecular weight cutoffs (MWCOs) and evaluated for the potential to support SC^neg NK92 cell growth. We decided to focus specifically on evaluating the potential of mbSC expression to confer effects mediated by mbSC components (rather than mbSC-expression-induced secretion of other factors by NK92 cells), and for that reason we evaluated media conditioned by transiently transfected HEK293FT cells (which should not express receptors for these factors).

To deplete EVs, conditioned media were ultrafiltered through a 100 kDa MWCO filter, which is sufficient to remove EVs.⁵⁷ To assess whether a lower limit in the size of bioactive compounds could be identified, media were then passed through a 50 kDa MWCO filter and a 10 kDa MWCO filter, which would remove free IL2 (∼16 kDa) and IL15 (∼15 kDa) (Fig. 7C). For both mbSCs, non-EV factors were sufficient to induce SC^neg NK92 cell growth, and these factors were greater than 10 kDa in size (Fig. 7E).

These factors may be mbSC cleavage products, as the transmembrane domains of both mbIL15 and mbIL2 are known to be cleaved at the membrane.^58,59 Intriguingly, mbIL2-conditioned media was significantly less effective than was mbIL15 conditioned media after filtration through a 50 kDa MWCO filter (Fig. 7E). This may reflect the removal of bioactive soluble mbIL2 multimers by that filter.⁶⁰ In sum, these observations indicate that paracrine activity of mbSC expression can be mediated by EVs as well as secreted subunits of these transgene products, and these effects vary by mbSC identity.

Evaluating the potential of combined synthetic cytokine expression for engineering NK92 cells

NK92 cell-based therapies that combine the expression of a membrane-bound synthetic cytokine (mbSC) and soluble synthetic cytokine (sSC) may benefit from both potent mbSC autocrine activation and delivery of sSCs to modulate endogenous immune cells. However, sustained cytokine overexposure can also lead to NK cell dysfunction.⁶¹ Therefore, these tradeoffs must be evaluated to identify suitable combinations and even interaction effects. As a first step toward evaluating how exposure to signaling from both an mbSC and sSC impacts NK92 cell expansion and cytotoxicity, we employed a consortia model in which populations of mbSC-expressing NK92 cells and sSC-expressing NK92 cells were mixed and evaluated.

We tested how mixed consortia of SC-expressing NK92 cells modulate expansion and cytotoxicity. Consortia of NK92 cells were incubated for 3 days, after which each subpopulation's cell count was measured (Fig. 8A, B). NK92 cells expressing mbIL2 or mbIL15 did not exhibit altered growth when mixed into consortia with NK92 cells expressing Neo2/15 or eIL15 (Fig. 8A), and the converse was also true (Fig. 8B). NK92 cells expressing eIL12 or DR18 exhibited greater expansion when co-cultured with mbIL2- or mbIL15-expressing NK92 cells (Fig. 8B).

FIG. 8.

SC-engineered NK92 cell consortia enhance growth and cytotoxicity.

(A, B) Expansion of SC-expressing NK92 cells when in consortia. MbSC NK92 cells were co-cultured with sSC NK92 cells at a 1:1 ratio for 3 days, after which mbSC NK92 cells (miRFP720^pos) (A) and sSC NK92 cells (mNeonGreen^pos) (B) were counted by flow cytometry. Normalized count was calculated by dividing SC NK92 cell counts by the average SC NK92 cell count when co-cultured with BB NK92 cells; see Equation (2) (Materials and Methods section). Bar height indicates mean normalized count. Error bars represent standard deviation. n = 3.

(C) Cytotoxicity of SC-expressing NK92 cells when in consortia. sSC NK92 cells alone, or mixed 1:1 with mbSC NK92 cells, were co-incubated with fluorescent K562 cells at a ratio of 3.125:1 NK92:K562 for 20 h. Live K562 cells (mScarlet-I^pos) were counted by flow cytometry and used to calculate %K562 growth inhibition. Bar height indicates mean %K562 growth inhibition. Error bars indicate standard deviation. n = 3. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 (two-way ANOVA followed by Tukey's HSD). n = 3. See Supplementary Note 1 for full ANOVA results.

Similarly, mbIL15-expressing NK92 cells exhibited slightly (but significantly) increased expansion when cultured with sp^IgE3xFLAG-DR18 NK92 cells (Fig. 8A), which may reflect synergistic effects of the two cytokines on proliferation.⁶² Overall, NK92 cell expansion was not diminished by (and in some cases was increased by) exposure to multiple SCs. We observed a similar pattern in evaluations of cytotoxicity, in that mixing mbSC NK92 cells and sSC NK92 cells into consortia did not diminish the overall population cytotoxicity (Fig. 8C).

Notably, SC lines showed varying capacity to rescue population cytotoxicity when in consortia with SC^neg NK92 cells. In particular, mbIL2 and eIL15 NK92 cells fully rescued population cytotoxicity, as their cytotoxicity as pure populations did not differ significantly from their cytotoxicity when in consortia with SC^neg NK92 cells. Altogether, these consortia analyses suggested that combinatorial engineering of NK92 cells could be feasible.

Discussion

The findings presented in this study have several implications for engineering NK92 cells to express synthetic cytokines and will inform future efforts to design NK92 cell therapies, which may expand the availability of cell-based cancer therapies to more patients. Although prior investigations have evaluated the expression of some synthetic cytokines in NK92 cells to enhance their function, no prior analysis includes comparative evaluation across properties that are important for cell therapy production and performance.

After platform development focused on enabling or enhancing expression of SCs, our evaluation included properties that are useful for both cell therapy production (selection of engineered cells, expansion in cytokine-free media) and performance (cytotoxicity, including after exposure to irradiation and hypoxia). Finally, we generated a series of new insights into the mechanisms by which SC-expressing NK92 cells may function in paracrine fashion, and in which combinatorial SC effects may drive NK92 cell growth and performance.

This study generated multiple novel insights that are important to enabling or enhancing expression of SCs. We established, for the first time, that DR18 and single-chain IL12, the latter only when expressed at attenuated levels, can be expressed constitutively by NK92 cells to a degree sufficient to increase NK92 cell cytotoxicity. The feasibility of constitutively expressing DR18 and eIL12 was surprising, as prolonged exposure to their native analogs induces NK92 cell death.^39,40

It is possible that low-level IL12 expression leads to levels of IL12 signaling closer to those experienced in vivo or that IL12 exerts distinct NK92 cell responses as a function of dose. We also found that Neo2/15 is highly sensitive to C-terminal modification and shows appreciable autocrine activity despite lacking a known secretion tag, though addition of an IgE signal peptide substantially enhanced its effects. Similarly, DR18 showed appreciable autocrine activity despite lacking a secretion tag.

This is surprising for Neo2/15, as it is a fully synthetic amino acid sequence that lacks known signals for secretion, while this is less surprising for DR18, as its sequence is derived from IL18, which is secreted via an unconventional pathway that is dependent on plasma membrane permeability.⁶³

This study also showed, for first time, that SCs impart substantially different selective pressures on NK92 cells. It was unsurprising that mbSCs, but not SCs, positively selected for NK92 cells when they were cultured in cytokine-free media, as mbIL2 and mbIL15 were known to preferentially support transduced NK92 cell growth,^21,23 while soluble SCs were expected to support the growth of both transduced and non-transduced cells.

However, it was surprising that mbIL2 selected for lower expression loci than mbIL15. One possibility is that mbIL2 activates NK92 cell growth signaling more potently than does mbIL15, enabling the survival of NK92 cells with lower transgene expression. Another possibility is that mbIL2 is toxic above certain expression levels, negatively selecting cells with high transgene expression loci. Regardless, the different selection mediated by these mbSCs is particularly surprising as the IL2 and IL15 receptors only differ in their alpha subunits, which are not thought to contribute to cytokine signaling.

However, the contribution of IL15Rα to IL15 signaling is becoming more appreciated,⁶⁴ and it is possible that signaling from IL15Rα in mbIL15 is responsible for the different mbSC selection patterns. It is also possible that mbIL2 and mbIL15 complex with other signaling chains with different affinity, leading to differential signaling. Finally, it is possible that mbIL2 and mbIL15 have different stabilities, necessitating different rates of expression for the same level of surface expression. It was also surprising that Neo2/15 exerted negative selective pressures, which is likely due to exhaustion from high cytokine signaling or cell burden due to transgenic protein production.⁶⁵

This study also generated insights into the impacts of SC expression on NK92 cell performance. As expected, SCs analogous to IL2 or IL15 supported NK92 cell expansion, but not when challenged with hypoxia or irradiation. Excitingly, all SCs but mbIL2 retained effects on cytotoxicity in hypoxia, and all SCs retained their effects on cytotoxicity when irradiated. The discrepancy between growth and cytotoxicity might be due to the disruption of distinct signaling mechanisms that mediate growth but not cytotoxicity, or may reflect that SC enhancement of cytotoxicity, but not expansion, is the result of being primed with cytokine signaling, rather than cytokine signaling concurrent with cytolysis.

Regardless, the SCs increase the cytotoxicity of NK92 cells without mitigating their growth inhibition by irradiation. This observation mitigates concerns that these SCs may override this critical safety process or that irradiation induced DNA-damage may impair the effects of SCs on cytotoxicity, but this requires further evaluation of the possibility of rare exceptions to this pattern. The loss of mbIL2 effects on cytotoxicity in hypoxia was unexpected and might reflect the induction of a state that is more sensitive to hypoxia.

This study additionally revealed insights into the paracrine effects of mbIL2 and mbIL15. Both mbSCs had contact-independent paracrine effects on NK92 cell expansion. These findings are likely due, in part, to receptor chain cleavage and release of soluble cytokine-cytokine receptor complexes, which have been described for both IL2-IL2Rβ and IL15-IL15Rα complexes, the cytokine-cytokine receptor pairs used in mbIL2 and mbIL15, respectively.^59,60 Notably, these prior investigations, as well as our study of sub-EV mediators of mbSC paracrine effects, did not express cytokines in NK92 cells, and it remains possible that these findings do not translate to paracrine effects of mbSCs produced by NK92 cells (e.g., due to differential expression of proteases responsible for cytokine receptor cleavage).

We also discovered that EVs were a novel mechanism by which mbIL15 exerts paracrine effects, and that, surprisingly, this mechanism did not also apply to mbIL2. The EVs isolated from mbIL15, but not mbIL2, expressing NK92 cells were sufficient to induce NK92 cell growth. There are multiple possible explanations for how mbIL15 mediates effects via vesicles. It might be presented on the surface of vesicles and directly activate IL15 receptors on NK92 cells.

It is also possible that vesicles are endocytosed and receptors are trafficked to a suitable membrane location in the recipient cell, where they influence signaling. The unexpected difference in the role of EVs in mediating paracrine effects of mbIL2 versus mbIL15 may be due to differential localization within intracellular membrane compartments,^66,67 which might affect their relative propensity to be packaged into vesicles.

Finally, our consortium analyses indicate that signaling from membrane-bound cytokines and soluble synthetic cytokines can be combined without compromising NK92 cell expansion or cytotoxicity, and such strategies even enhance the expansion of NK92 cells expressing DR18 or eIL12. This finding supports the potential utility of developing NK92 cell therapies that combine such modalities. Combining modalities is an attractive design, because mbSCs enable the production and performance of engineered NK92 cell therapies, while sSCs have previously demonstrated abilities to activate antitumor responses by endogenous immune cells or co-administered adoptive cell therapies.^20,24–27

While the findings of our consortium analysis were largely expected, the outcomes were nonetheless non-obvious, as it was possible that cytokine overstimulation would lead to cell dysfunction or death. These analyses also provide a useful reference point for future efforts to engineer NK92 cells with a single vector that combines the expression of mbSCs with sSCs, which may be complicated by competing selective pressures, resource competition from high protein expression in a single cell, or the effects of prolonged exposure to more than one cytokine.

A few key considerations guide the interpretation of this study and identify opportunities for future investigation. First, generating SC-expressing NK92 cell lines involved the selection and expansion of sub-populations of NK92 cells, and such lines may show some random variation in traits other than SC expression. This may be addressed by assessing variation in the behavior across multiple NK92 cell lines that express the same SC.

Second, we did not directly quantify the levels of SC expression or secretion by NK92 cells, and in some cases, doing so requires the development and validation of new physical assays (e.g., enzyme-linked immunosorbent assays) that are specific for these SCs. Such quantification would be particularly important for comparing the therapeutic benefit of SCs alone (e.g., as recombinant proteins) versus the administration of SC-expressing cells.

Finally, it is known whether the phenomena demonstrated on NK92 cells extend to related cell types, most notably primary NK cells and induced pluripotent stem cell-derived NK cells. Although NK92 cells mirror the phenotype of activated primary NK cells, these cell types differ in substantial ways, and such a comparison is an exciting avenue for future research. We hope that the findings described herein will provide a useful reference point and source of hypotheses for engineering primary NK cells. Such comparisons may ultimately provide mechanistic insights into the impact of SC-expression on NK92 cells versus primary NK cells.

Finally, we consider prospects for translating the engineering strategies reported here to clinical evaluation and impact. An immediate next step is preclinical evaluation of efficacy and safety in appropriate models (e.g., humanized mice with xenografted human tumors). These studies should evaluate the anti-tumor efficacy of SC-expressing NK92 cells, with careful attention paid to the evaluation of safety, as systemic administration of related natural cytokines leads to toxicity above certain thresholds.

The SC expression should also be investigated in cell types with demonstrated clinical utility, most notably including CAR-expressing primary NK cells, which have proven curative in a Phase I/II clinical trial.⁸ Overall, the findings described herein contributed multiple new biological insights into the behavior and effects of SCs expressed in NK92 cells, which will enable future engineering efforts that leverage SCs to enhance cell therapy production and performance.

Conclusions

This comparative analysis of the effects and interactions of synthetic cytokine expression may guide NK92 cell bioengineering and improve the feasibility and function of NK92 cell-based therapies. This study generated a suite of hypotheses that may guide future validation work relating in vitro performance to in vivo performance, which is most appropriately deployed across a range of cancer models, as performance varies by model and disease.

Another frontier is the combination of these synthetic cytokine strategies with other functional transgenes (e.g., expression of CARs to enhance tumor-reactivity). While this study focused on NK92 cell therapies, the framework for comparatively evaluating synthetic cytokines may be extended to evaluate other cell-based therapies. Ultimately, this analysis revealed multiple avenues for continuing to improve NK92 cell bioengineering toward realizing the potential of effective, off-the-shelf anti-cancer cell therapeutics.

Materials and Methods

General DNA assembly and synthetic cytokine modification

Plasmid construction was completed using standard molecular biology techniques. Polymerase chain reactions were completed using Phusion DNA Polymerase (New England Biolabs). Plasmids were assembled using restriction enzyme cloning. Transgenes were codon-optimized for production in human cells, using the GeneArt Optimizer (ThermoFisher). A (GGGGS)₃ linker was inserted between the two IL12 monomers to enhance protein stability.⁶⁸ A 3xFLAG tag followed by a GSG linker was added to the N terminus of Neo2/15.

A 3xFLAG tag followed by a GSG linker was added to the N terminus of DR18. The FLAG epitope tag was removed between the signal peptide and coding sequences of eIL15. Plasmids created in this study are deposited with and distributed by Addgene at (https://www.addgene.org/Joshua_Leonard/). Plasmid sequences for constructs created in this study are included in Supplementary Data.

Cell culture

HEK293FT (R70007; Thermo Fisher) and HEK293T Lenti-X (#632180; Takara Bio) cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (#31600-091; Gibco) supplemented with 10% heat-inactivated fetal bovine serum (#16140-071; Gibco), 100 U/mL penicillin, 100 μg/mL streptomycin (#15140122; Gibco), and 4mM L-glutamine (#25030-081; Gibco). For Lenti-X cells, medium was supplemented with 1 mM sodium pyruvate (#11360070; Gibco).

To split HEK293FT and Lenti-X cells, spent medium was aspirated, cells were rinsed with 5 mL of sterile phosphate buffered saline (PBS) and then incubated with 1.5 mL of Trypsin-EDTA at 37°C for 2–5 min. Cells were resuspended in fresh media, then plated in 10 cm dishes. K562 B cell lymphoma cells (CCL-243; ATCC) were cultured in Iscove's Modified Dulbecco's Medium (IMDM) (#12200036; Gibco) supplemented with 10% heat-inactivated fetal bovine serum and 100 U/mL penicillin, 100 ug/mL streptomycin.

The Bigger Picture

Off-the-shelf immune cell therapies are an increasingly promising strategy for increasing patient access to powerful emerging modalities of cancer treatment. A key open challenge is matching the activity and efficacy of primary autologous or allogeneic cell products with cells derived from immortalized sources. In this study, we systemically evaluated the use of synthetic cytokine expression to potentiate NK-like NK92 cells, and we identified cell engineering strategies that improve expansion and cytotoxic activity.

These findings provide a foundation for subsequent preclinical development. These reagents and comparisons provide a framework for investigating the use of synthetic cytokines in other cell-based therapies, including those derived from primary NK cells or T cells and those that integrate other technologies such as CAR expression. We hope that such bioengineering analyses will continue to improve the safety, efficacy, and availability of curative cell therapies.

K562 cells were either subcultured at a 1:10 or 1:20 ratio or resuspended at 1E5 cells/mL every 2–3 days. NK92 cells (CRL-2407; ATCC) were cultured in minimal essential medium (MEM) α (12000-022; Gibco) with 0.2 mM Myo-inositol (I-7508; Sigma), 0.1 mM 2-mercaptoethanol (21985-023; Gibco), 0.02 mM folic acid (F-8758; Sigma), 1.5 g/L sodium bicarbonate, 12.5% fetal bovine serum (16000044; Gibco), 12.5% horse serum (16050122; Gibco), and 100 U/mL penicillin, 100 μg/mL streptomycin.

To culture, cells were kept in T25 or T75 flasks (431464U, 431463; Corning) and maintained between a density of 2 × 10⁵ cells/mL and 1 × 10⁶ cells/mL. Medium was supplemented with 100 IU/mL recombinant IL2 (200-02; Peprotech) for parental (unmodified), mbSC backbone (BB), sSC BB, sp^null3xFLAG-Neo2/15, eIL12, sp^null3xFLAG-DR18, and sp^IgE3xFLAG-DR18 NK92 cells.

Calcium phosphate transfection

Plasmid DNA in nuclease-free water was mixed with 2 M CaCl₂ (final concentration 0.3 M), pipetted eight times to mix, then added dropwise to an equal volume of 2 × 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid (HEPES) buffered saline (280 mM NaCl, 50 mM HEPES, 1.5 mM Na₂HPO₄, pH 7.1), and pipetted up and down four times to mix. After 3 min and 30 s, the mixture was pipetted vigorously eight times, then added dropwise to cells. Cell medium was changed to fresh medium after 8–16 h.

SC conditioned medium production

The 4.5 × 10⁶ HEK293FT cells were plated in 10 cm dishes and allowed to adhere for 8 h. The cells were transfected with 20 μg of cytokine plasmid DNA and 1 μg of a plasmid encoding dsRedExpress2 using the calcium phosphate method. Twenty-four hours after medium change, the supernatant was removed, spun 125 g 5 min at 4°C, and then filtered through a 0.45-μm filter. Medium was used within a week of harvest.

NK92 cell line generation

The 5 × 10⁶ HEK293FT cells were allowed to adhere for 8 h in 10 cm dishes, then transfected with 8 μg psPAX2 (12260; Addgene), 3 μg PMD2.G (12259; Addgene), 10 μg transfer plasmid, and 500 ng plasmid encoding dsRedExpress2⁶⁹ as a transfection control. Two 10 cm dishes were transfected per lentivirus type. The next morning, the medium was changed to 10 mL fresh DMEM and plates were incubated for 28 h. Lentivirus-containing supernatant was pulled up from plates, centrifuged at 500 g for 2 min at 4°C, and passed through a 0.45-μm filter.

Four milliliters of NK92 cells (1 × 10⁶ cells/mL in complete MEM α) was added to the lentiviral supernatant, along with 4 μg/mL of polybrene in 0.9% saline (TR-1003-G; EMD Millipore). Cells were spun in viral supernatant at 37°C for 90 min at 2500 g in a fixed angle rotor, then resuspended in viral supernatant, and incubated overnight. The next morning, cells were spun at 125 g for 5 min at 4°C, viral supernatant was aspirated, and cells were resuspended in fresh media. Transduction was analyzed 2 days later (BD LSRFortessa SORP Cell Analyzer). To select for cells that expressed sSCs, cells were cultured with puromycin (1 μg/mL), then enriched by FACS (BD FACSAria IIu Cell Sorter).

NK92 cells transduced with the mbSC BB were selected with blasticidin (10 μg/mL) for 14 days. NK92 cells transduced with mbIL2 were selected via culture without IL2 followed by selection with blasticidin (10 μg/mL) for 14 days. NK92 cells transduced with mbIL15 were selected via culture without IL2.

K562 cell transduction

The 5 × 10⁶ Lenti-X HEK293T cells were plated in 10 cm dishes and allowed to adhere for 24 h. Cells were transfected with 8 μg psPAX2 (12260; Addgene), 3 μg PMD2.G (12259; Addgene), and 10 μg transfer plasmid, using the calcium phosphate method. One microgram of plasmid encoding dsRedExpress 2 was cotransfected as a transfection control.⁶⁹ The next morning, media were replaced with fresh medium. Cells were incubated for 32 h. Lentivirus-containing supernatant was collected from plates, centrifuged at 500 g for 2 min at 4°C, and passed through a 0.45-μm filter. One milliliter of virus was added to 100 μL of K562 (1 × 10⁶ cells/mL) in a 12-well plate. Transduction was analyzed 3 days later (BD LSRFortessa SORP Cell Analyzer).

Western blot analysis of protein expression

For analysis of EV markers, NK92 cells were seeded at 4 × 10⁵ cells/mL in incomplete MEM α and after 2 days, culture was centrifuged at 125 g 5 min to separate NK92 cells and conditioned medium. Vesicles were isolated from conditioned medium as described next but, after size exclusion chromatography, were concentrated with filters that had not been treated with bovine serum albumin. To obtain NK92 cell lysate, NK92 cells were rinsed with PBS and incubated with radioimmunoprecipitation assay buffer (RIPA buffer; 150 mM NaCl, 50 mM Tris-HCl pH 8.0, 1% Triton X-100, 05% sodium deoxycholate, 0.1% sodium dodecyl sulfate) with protease inhibitor cocktail tablet (PIA32953; Pierce; 1 tablet per 10 mL RIPA buffer) for 5 min at room temperature (∼20–25°C), incubated on ice for 30 min, and then spun at 12,200 g for 20 min to clear supernatant, which was used in subsequent analyses.

To produce cell lysates for synthetic cytokine analysis, 4.5 × 10⁶ HEK293FT cells were transiently transfected with 20 μg of cytokine plasmid DNA using calcium phosphate transfection. Thirty-six hours after media change, cell lysate was harvested as done for NK92 cells. Cell lysate protein content was analyzed using a bicinchoninic acid assay (Pierce).

Lysate or vesicles were incubated with either reducing or non-reducing Laemmli buffer at either 70°C or 90°C for 10 min (Table 1). For synthetic cytokine expression analysis, 5 μg of protein per well was loaded into a 4–15% polyacrylamide gradient Mini-PROTEAN TGX precast protein gel (Bio-Rad). For vesicle marker analysis, either 3 μg of protein from cell lysates or 4.8 × 10⁸ vesicles per well were loaded. Gels were run at 50 V for 10 min, then 100 V for 60 min at room temperature. Protein was then transferred to a polyvinylidene difluoride membrane (Bio-Rad) at 100 V for 45 min.

Table 1.

Antibody properties and sample preparation methods used in western blot analysis

Antibody target	Product no.	Laemmli buffer	Denature temperature (°C)	Antibody dilution	Animal of origin
3xFLAG	Sigma F1804	Reducing	70	1:1000	Mouse
Myc	Abcam ab32	Reducing	70	1:1000	Mouse
IL12p40	Abcam ab106270	Reducing	70	1:1000	Rabbit
IL15	Abcam ab7213	Reducing	70	1:1000	Rabbit
Calnexin	Abcam ab22595	Reducing	70	1:1000	Rabbit
CD81	Santa Cruz sc23962	Non-reducing	90	1:500	Mouse
Mouse	Cell Signaling 7076	N/A	N/A	1:3000	Horse
Rabbit	Invitrogen 32460	N/A	N/A	1:3000	Goat

Membranes were blocked in either 3% non-fat milk in tris buffered saline (TBS; 50 mM Tris, 138 mM NaCl, 2.7 mM KCl, pH 8.0) for 30 m at room temperature (3xFLAG) or 5% non-fat milk in TBS plus tween (TBST; 50 mM Tris, 150 mM NaCl, 0.1% Tween 20, pH 7.6) for either 1 h at room temperature (Myc, Calnexin, CD81) or overnight at 4°C (IL12p40, eIL15), then incubated with primary antibody (Table 1). For 3xFLAG immunoblotting only, membranes were rinsed 3 × 5 min with TBS, with 0.05% Tween 20 added in the second and third rinse, before the addition of primary antibody.

After incubation with the primary antibody, all membranes were rinsed 3 × 5 min with TBST, then incubated with secondary antibody for 1 h room temperature. Afterward, the membranes were rinsed 3 × 5 min in TBST before the addition of Clarity Western ECL reagent (BioRad) and subsequent exposure to film or chemiluminescence imaging with an Azure c280 imager.

NK92 cell irradiation

NK92 cells resuspended in MEM α were irradiated with 10 Gy of X-ray irradiation at a rate of 208 cGy/min, using a RadSource RS-2000 irradiator set to 160 kV and 25 mA, then immediately used in assays.

Flow cytometry-based cytotoxicity assays

NK92 cells were incubated without IL2 for 1 day, then resuspended in MEM α at the maximum effector concentration tested, and serially diluted twofold to make cell suspensions for lower effector concentrations. One hundred microliters of NK92 cells in MEM α, or MEM α alone, were added to 100 μL of mScarlet-I expressing K562 resuspended in IMDM at 1 × 10⁵ cells/mL, in a 96-well plate.⁷⁰ Cells were incubated for 20 h at 37°C, 5% CO₂, and either ambient oxygen (∼21% O₂) or hypoxia (1% O₂), using a hypoxic incubator (HERAcell150i; Thermo Fisher).

Three μM 4′,6-diamidino-2-phenylindole (DAPI) and PKH26 reference microbeads (P7458; Sigma) were added to each sample, and samples were analyzed using a BD LSRFortessa SORP Cell Analyzer (Supplementary Fig. S2A). Live K562 cells (DAPI^neg, mScarlet-I^pos) counts were analyzed and normalized to the number of reference microbeads counted, to obtain a count of live tumor cells per well. For cytotoxicity assays, percent growth inhibition was defined as follows:

$$Percentgrowthinhibition=100\times \left(1-\frac{liveK562s^{+NK92s}}{averageliveK562s^{-NK92s}}\right)$$ (1)

Flow cytometry-based growth assays

Unirradiated NK92 cells, or NK92 cells irradiated as described earlier, were resuspended at 2 × 10⁵ cells/mL. Cells were plated in two sets of 96-well plates (200 μL/well), then incubated at 37°C, 5% CO₂ and either ambient O₂ or 1% O₂. After 2 days or 4 days, either 3 μM DAPI or 0.3 μg/mL propidium iodide, and PKH26 reference microbeads (P7458; Sigma) were added to one set of plates, and each sample was analyzed using a BD LSRFortessa SORP Cell Analyzer (Supplementary Fig. S2A).

The number of live NK92 cells detected was normalized to the number of reference microbeads detected to obtain live NK92 cell counts per well. For consortia cell expansion assays, normalized cell count was calculated with the following formula:

$$Normalizedcount=\frac{liveSCNK92s^{experimentalcondition}}{averageliveSCNK92s^{consortiawithBBNK92s}}$$ (2)

Time lapse microscopy killing assays

NK92 cells were incubated for 1 day without IL2. K562 transduced to express mScarlet-I were resuspended in IMDM at 1 × 10⁵ cells/mL. For 4 h videos of killing, 3 μM DAPI was added to each well and wells were imaged every 5 min for 4 h total with a 10 × objective, an ET-dsRed filter (49005; Chroma), a DAPI filter (OP-87762; Keyence), and either a Cy5.5 filter (49022; Chroma) to detect mbSC NK92 cells or an EYFP filter (49003; Chroma) to detect sSC NK92 cells.

One hundred microliters of K562 cells per well were plated in a 96-well plate. NK92 cells were resuspended in MEM α at 3.125 × 10⁵ cells/mL and 100 μL of NK92 cells were added to K562 cells, for a total volume of 200 μL. For 20 h killing kinetic assay, the plate was imaged with a 10 × objective and an ET-dsRed filter every 2 h for 20 h total at five points per well in a time-lapse microscope (BZ-X800; Keyence) fitted with an incubation chamber (Tokai Hit). Image-based cell-cytometry software (Keyence) was used to calculate K562 cell counts.

EV isolation

NK92 cells plated at 4 × 10⁵ cells/mL were cultured for 2 days in 10 mL of incomplete MEM α. Conditioned medium was centrifuged at 300 g for 10 min, and supernatant was further centrifuged at 2000 g for 20 min to remove dead cells and apoptotic bodies (J-LITE JLA 16.25 rotor; Beckman Coulter Avanti J-26XP centrifuge). Supernatant was centrifuged at 26,500 rpm for 2 h 21 min (SW41 Ti rotor; Beckman Coulter Optima L-80 XP ultracentrifuge), using polypropylene ultracentrifuge tubes (331372; Beckman Coulter).

All centrifugation steps were performed at 4°C. Supernatant was aspirated until ∼100 μL remained, and EV pellet was left on ice for 30 min. Resuspended EVs were transferred to microcentrifuge tubes, then run on a size exclusion chromatography column (ICO-70; Izon), using PBS as a running buffer. The first 2 mL of eluent was collected and re-concentrated using 50 kDa ultrafilters (UFC8050; Amicon) that been precoated with 1% bovine serum albumin in PBS for 1 h at room temperature; eluent was centrifuged for 30 min at 4000 g, 4°C to concentrate the EVs.

Nanoparticle tracking analysis

Vesicle concentration and size were measured using a Nanosight NS300 (Malvern) running software v3.4 and a 642 nm laser. Vesicles were diluted to 2–10 × 10⁸ particles/mL in PBS for analysis. Samples were run at an injection rate of 30, imaged at a camera level of 14, and analyzed at a detection threshold of 7. Three 30-s videos were captured for each sample to determine the average vesicle concentration and size histograms.

Transmission electron microscopy

Ten microliters of purified EVs were placed onto a carbon-coated copper grid (CF400-Cu-50; Electron Microscopy Services) for 10 min before excess liquid was wicked away with a piece of filter paper. The grid was dipped in PBS twice to remove excess proteins and unreacted ligands from the media and reaction, and it was allowed to dry for 2 min. To achieve negative staining, 10 μL of uranyl acetate solution (2 wt% in Milli-Q water) was placed on the grid for 1 min before being wicked away with filter paper.

The grid was allowed to fully dry (3 h to overnight) at room temperature (∼20°C). Bright-field transmission electron microscopy (TEM) imaging was performed on a JEOL 1230 TEM. The TEM operated at an acceleration voltage of 100 kV. All TEM images were recorded by a Hamamatsu ORCA side-mounted camera or a Gatan 831 bottom-mounted CCD camera, using AMT imaging software.

Serial filtration of mbSC-conditioned media

Transfected HEK293FT cells conditioned media for 28 h. Cells were removed from medium by centrifugation at 125 g for 5 min, then filtration through a 0.45-μm filter. Conditioned media were serial filtered by first centrifuging at 4000 g for 20 min in a 100 kDa MWCO filter (UFC9100; Amicon), then centrifuging at 4000 g for 10 min in a 50 kDa filter (UFC9050; Amicon) and finally, centrifuging at 4000 g for 30 min for 10 kDa device (UFC9010; Amicon). All centrifugation was performed at 4°C.

Supplementary Material Online

Online materials include Supplementary Information, Supplementary Videos (see Supplementary Note 3 for descriptions), plasmid sequence files for constructs generated in this study, and source data including flow cytometry files, raw images, and data used to generate figures.

Authors' Contributions

S.D. designed and performed experiments, as well as prepared the manuscript. P.S.D., J.A., and B.Z. contributed to experimental design. R.E.M. performed experiments. I.J.H.-K. aided in the execution of experiments. J.N.L. contributed to experimental design and manuscript preparation.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was funded by in part by an AbbVie pilot project, a Lurie innovation award, and the National Institute of Biomedical Imaging and Bioengineering of the NIH under award number 1R01EB026510. This work was supported by the Northwestern University Center for Genetic Medicine and Robert H. Lurie Comprehensive Cancer Center Flow Cytometry Core Facility. The Lurie Cancer Center is supported in part by an NCI Cancer Center Support Grant no. P30 CA060553.

Supplementary Materials

Supplementary Data

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS1.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS2.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS3.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS4.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS5.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS6.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS7.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS8.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS9.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS10.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS11.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS12.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS13.mp4

/doi/suppl/10.1089/genbio.2023.0024/suppl_file/Supp_VideoS14.mp4