The Activity of EGFR CAR NK and CAR T Cells against EGFR Inhibitor-Resistant NSCLC and Drug-Tolerant Persister Cells.

INTRODUCTION
Non-small cell lung cancer (NSCLC) is the leading cause of cancer-related deaths worldwide. About 15% of NSCLC cases harbor epidermal growth factor receptor (EGFR) activating mutations, and EGFR tyrosine kinase inhibitors (TKIs) such as osimertinib provide significant clinical benefit for patients, with a median progression-free survival (PFS) of more than 18 months observed for osimertinib monotherapy and a longer PFS for combination regimens(1–4). Despite initial clinical benefit, however, in virtually all patients, a population of residual drug-tolerant persister cells (DTPCs) is not eliminated by TKI therapy and eventually evolves to become drug-resistant cells (DRCs), resulting in disease progression. Mechanisms mediating resistance to the EGFR TKI osimertinib are highly heterogeneous, encompassing EGFR-dependent mechanisms such as the acquisition of additional EGFR mutations, and EGFR-independent mechanisms such as the gained expression of other cancer-driver genes(5,6) and cell lineage shifts such as epithelial to mesenchymal transition (EMT) (7–11) and SCLC transformation(12). Moreover, EGFR mutant NSCLC cells with EMT-mediated resistance also acquire broad-spectrum drug resistance, highlighting the clinical challenge associated with treating this patient population (11). In addition to the development of targeted agents, the treatment of NSCLC has been revolutionized by immunotherapy. Recent clinical studies treating NSCLC patients with immune checkpoint inhibitors (ICIs) have demonstrated significant improvements in clinical outcomes (13–15). However, EGFR-TKI resistant NSCLC patients are typically poorly responsive to ICIs(16–21).
Chimeric antigen receptor (CAR)-based cell therapy is an attractive therapeutic approach that involves the engineering of immune cells, such as T cells and NK cells, to express CARs to attack the tumor cells by recognizing a specific target on the cell surface. CAR-T cell therapies have shown promising results for the treatment of B cell-originated hematological malignancies, and the FDA has approved six CAR-T cell therapies since 2017. Although CAR-T cells have shown promise in solid tumors, there are still challenges, such as a limited number of appropriate cell-surface antigens and the immunosuppressive tumor microenvironment (TME)(22–24). Moreover, the expression of target antigen in normal tissues, even at low levels, can also lead to on-target/off-tumor toxicity of CAR-T cells. For example, a case report showed a serious adverse event following infusion of a large number of CAR-T cells with high affinity to HER2(25). CAR-NK cell therapies have been explored more recently. While they have the potential disadvantage of potentially shorter in vivo persistence, they also have important potential advantages, including a more favorable safety profile, “off-the-shelf” potential, and antibody-dependent cellular cytotoxicity (ADCC), and have demonstrated promising early results (26–30). Thus, CAR-NK cell therapy could be an alternative to CAR-T cell approaches.
We hypothesized that EGFR-targeting cellular immunotherapy would be an effective approach for treating NSCLC cells with acquired EGFR TKI-resistance as well as EGFR DTPCs. To test this, we generated EGFR CAR-T and CAR-NK cells and investigated their therapeutic efficacy in osimertinib-resistant models in vitro and in vivo. We observed that NSCLC cells with acquired resistance to EGFR TKIs had increased sensitivity to EGFR CAR-NK cells as compared to treatment-naïve parental cells. We also investigated mechanisms by which EGFR TKI resistant cells promote an immunosuppressive tumor microenvironment and demonstrate that NSCLC cells with acquired EGFR TKI resistance markedly upregulate expression of the immunosuppressive cytokine TGF-β. Moreover, we show that the antitumor activity of EGFR CAR-NK cells can be enhanced by blockade of TGF-β signaling. Our results support future investigations of EGFR CAR-NK cells for treating NSCLC patients with acquired EGFR TKI-resistance and DTPCs.

MATERIALS AND METHODS

Materials
K-562 (RRID:CVCL_0004) and 293T (RRID:CVCL_0063) cell lines were purchased from ATCC. HCC827 (RRID:CVCL_2063), H1975 (RRID:CVCL_1511), and HCC4006 (RRID:CVCL_1269) were obtained from Dr. John Minna (UT Southwestern). All cell lines were authenticated by STR profiling. K-562 (RRID:CVCL_0004), HCC827 (RRID:CVCL_2063), H1975 (RRID:CVCL_1511), and HCC4006 (RRID:CVCL_1269) were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS), 1% penicillin-streptomycin, and 1% L-glutamine. 293T cells (RRID:CVCL_0063) were cultured in DMEM medium supplemented with 10% FBS, 1% penicillin-streptomycin, and 1% L-glutamine. EGFR TKI resistant HCC827, H1975, and HCC4006 cells were established as previously described(11,31). All cell lines were tested negative for mycoplasma contamination.
Recombinant human IL-2 (200–02), IL-15 (200–15), TGF-β1 (100–21), and TGF-β2 (100–35B) were obtained from Peprotech. AF647-conjugated anti-IL-21 (RRID:AB_1227662), PE/Cy7-conjugated anti-4–1BBL (RRID:AB_2783179), PE-conjugated anti-OX40L (RRID:AB_3106062), BV421-conjugated anti-CD86 (RRID:AB_10899582), APC-conjugated anti-EGFR (RRID:AB_3675106), FITC-conjugated anti-CD3 (RRID:AB_571907), PE-conjugated anti-CD56 (RRID:AB_604101), FITC-conjugated anti-Granzyme B (RRID:AB_2687029), APC/Cyanine7-conjugated anti-human CD45 (RRID:AB_314402), PerCP/Cyanine5.5-conjugated anti-human CD4 (RRID:AB_893328), PE/Cyanine7 anti-human CD8 (RRID:AB_314117), FITC-conjugated anti-human PD-1 (RRID:AB_940477), APC/Cyanine7-conjugated anti-human TIM3 (RRID:AB_2565716), PerCP/Cyanine5.5-conjugated anti-human LAG-3 (RRID:AB_2910413), PE-conjugated anti-human TIGIT (RRID:AB_2632730), APC-conjugated anti-Perforin (RRID:AB_2571968), anti-NKp30 (RRID:AB_2814183), anti-NKG2D (RRID:AB_2561488), and mouse IgG1 (RRID:AB_11146992) were purchased from BioLegend. PE-conjugated anti-Phospho-Smad2 (Ser465/Ser467) (RRID:AB_2798933) was purchased from Cell Signaling Technology; Anti-human CD3/CD28 Dynabeads (11131D; RRID:AB_3676329) were purchased from Thermofisher. AF647-labled F(ab’)2 fragment of goat anti-human IgG F(ab’)2 (RRID:AB_2337898), AF647-labled F(ab’)2 fragment of goat anti-mouse IgG F(ab’)2 (RRID:AB_2338928), AF647-labled F(ab’)2 fragment of goat anti-human IgG (H+L) (RRID:AB_2337897), and AF647-labled F(ab’)2 fragment of goat control IgG (RRID:AB_2337017) were obtained from Jackson ImmunoResearch.

CAR construction
The second-generation 8BB3z CAR contains a single-chain variable fragment (scFv), a CD8 spacer and transmembrane domain, a 4–1BB intracellular domain, and a CD3ζ intracellular domain. The third generation 828BB3z CAR contains a scFv, CD8 spacer and transmembrane domain, a CD28 intracellular domain, a 4–1BB intracellular domain, and a CD3ζ intracellular domain. The third-generation M228BB3z CAR contains a scFv, CH2 and CH3 domains of IgG2 with N297Q mutation, CD28 transmembrane and intracellular domains, a 4–1BB intracellular domain, and a CD3ζ intracellular domain. scFvs derived from EGFR-specific antibodies, including cetuximab (CTB) and panitumumab (PNB), were used to generate EGFR-targeting CARs.
Genes encoding CARs were synthesized as gene fragments (Genewiz) and then cloned into an SFG retroviral expression vector (RRID:Addgene_22493; a gift from Martin Pule). Retroviral supernatant was produced by co-transfecting 293T cells (RRID:CVCL_0063) with plasmids encoding the CARs, the RD114 envelope protein, and packaging proteins (gag and pol). Retroviral supernatant was harvested and filtered through 0.45 μm filters after 48–72 hours post-transfection.

Generation of EGFR CAR-NK and EGFR CAR-T cells
To generate CAR-T cells, peripheral blood mononuclear cells (PBMCs) were isolated from the blood of healthy donors via Ficoll-Paque density gradient centrifugation. Isolated PBMCs were resuspended in complete RPMI-1640 media supplemented with 10% FBS, 2 mM GlutaMAX, 1% penicillin-streptomycin. To activate the T cells, anti-CD3/CD28 Dynabeads (RRID:AB_3676329) were added to PBMCs at a beads-to-cell ratio of 1:1 and incubated at 37°C in an incubator with 5% CO2. On day 3 post-activation, the anti-CD3/CD28 Dynabeads were removed magnetically, and the cells were transduced with a CAR-encoding retroviral supernatant. For transduction, non–tissue culture–treated 24-well plates were pre-coated with RetroNectin (20 μg/mL, Takara Bio) overnight at 4°C. After coating, retroviral supernatant was added, and the plates were centrifuged at 3,000 × g for 2 hours at 32°C to facilitate virus binding. The viral supernatant was then removed, and the activated T cells were added to the RetroNectin-coated plates. Two days after transduction, transduced CAR-T cells were transferred to cell culture flasks and cultured in fresh complete RPMI-1640 medium supplemented with 100 U/mL recombinant human IL-2 (PeproTech) at a density of 0.5 – 1 × 106 cells/mL.
To generate CAR-NK cells, a feeder cell line was first established by retrovirally transducing the human K562 cell line (RRID:CVCL_0004) with retroviral vectors encoding membrane-bound IL-21, 4–1BBL, OX40L, and CD86, respectively, to support NK cell activation and expansion. Successfully transduced cells were sorted by flow cytometry and expanded, then lethally irradiated with 100 gray (Gy) γ-rays prior to use as feeder cells. For NK cell activation, PBMCs from healthy donors were co-cultured with the irradiated feeder cells at a PBMC:Feeder ratio of 1:2 in RPMI-1640 complete media with additional cytokines (200 U/mL IL-2 and 5 ng/mL IL-15) to support NK cell proliferation and activation. On day 4 post-activation, enriched NK cells were transduced with retroviral supernatant on plates coated with RetroNectin (Takara) using the same procedure described for CAR-T cell transduction. Two days after transduction, CAR-NK cells were transferred to cell culture flasks and maintained in complete media supplemented with 200 U/mL IL-2 at 0.5 – 1 × 106 cells/ml.
All CAR-T and CAR-NK cells were used after they reached the resting state, which is about 7 days after transduction, and was confirmed by cell size on a flow cytometer.

Flow cytometry analysis
For cell surface staining, cell samples were collected, washed with PBS, stained with fluorescence conjugated antibodies in FACS staining buffer (PBS with 1% FBS) on ice for 30 minutes. The expression of CAR on CAR-T or CAR-NK cells was assessed by staining with an anti-mouse IgG F(ab’)2 antibody (RRID:AB_2338928) to detect scFv of CTB and using an anti-human IgG F(ab’)2 antibody (RRID:AB_2337898) to detect scFv of PNB. The expression of M228BB3z CAR was also analyzed by staining with an anti-human IgG (H+L) antibody (RRID:AB_2337897) to detect the IgG2 CH2 and CH3 domains in the CAR. To detect Granzyme B and perforin, NK cells were fixed using Fixation Buffer (Biolegend) at room temperature for 20 minutes. The cells were then permeabilized and stained with fluorescence conjugated anti-Granzyme B (RRID:AB_2687029) and perforin (RRID:AB_2571968) antibodies in Intracellular Staining Perm Wash Buffer for 20 minutes in the dark at room temperature. To detect phosphorylation of SMAD2 (p-SMAD2), NK cells were fixed with Fixation Buffer (Biolegend) at 37°C for 15 minutes at the end of the incubation with TGF-β. Samples were then permeabilized by adding pre-chilled True-Phos Perm Buffer (Biolegend) and incubating at −20°C overnight. Next, samples were stained with PE-conjugated anti-pSMAD2 antibody (RRID:AB_2798933) in FACS staining buffer at room temperature for 30 minutes. Samples were acquired on a FACSCanto (BD Biosciences), Accuri C6 (BD Biosciences), or Gallios (Beckman Coulter) flow cytometer. Data were analyzed using a FlowJo software (RRID:SCR_008520).

Cytotoxicity assays with CAR-NK and CAR-T cells
To quantify the killing activity of EGFR CAR-T and CAR-NK cells in vitro, luciferase (Luc) reporter assays were performed. For Luc reporter assays, the tumor cell lines were transduced with pHIV-Luc-ZsGreen (RRID:Addgene_39196; a gift from Bryan Welm) lentivirus to generate LUC-stable-expression cell lines. Target cells (Luc-expressing tumor cells) were seeded into a 96-well white cell culture plate (Thermo Fisher Scientific) at 1 × 104 cells in 100 μL complete media overnight. Effector cells (CAR-T and CAR-NK cells) were added to each well based on the indicated effector:target (E:T) ratio. The plate was then incubated in a 37 °C incubator with 5% CO2 for 4 hours. The supernatant was gently discarded after the incubation, and 100 μL of D-luciferin at 150 μg/ml was added. Luminescence was read using a FLUOstar OPTIMA multi-mode microplate reader (BMG Labtech). The percentage of specific lysis was calculated from luminescence as follows: ([target only – experiment]/[target only – no target] × 100).
For long-term killing assays, target cells were seeded into a 24-well cell culture plate (Thermo Fisher Scientific) at 1 × 105 cells in 500 μL of complete media. After 24 hours, CAR-T cells were added to each well at 5 × 105 cells in 500 μL of complete media. The plate was incubated in a 37 °C incubator with 5% CO2 for 48 hours. At the end of the incubation, cells were collected for flow cytometry to assess the expression of exhaustion-associated markers.

TGF-β blockade assays
For TGF-β blockade assays, the target cells were initially pre-seeded overnight as described above. The supernatant was then carefully removed, and media containing either 5μM galunisertib or DMSO were added into the wells of the indicated group. Subsequently, an equal quantity of effector cells (E:T = 1:1) was added to each well. The plates were then incubated in a 37 °C incubator with 5% CO2 for 72 hours.

NKp30 and NKG2D blockade assays
For NKp30 and NKG2D blockade assays, control NK or EGFR-CAR NK cells were incubated on ice for 30 minutes with 20 ug/ml of the indicated antibodies, including isotype control (RRID:AB_11146992) anti-NKp30 (RRID:AB_2814183), anti-NKG2D (RRID:AB_2561488), or a combination of both anti-NKp30 and anti-NKG2D. Following the incubation, these treated effector cells were added into 96-well plates that were previously seeded with indicated tumor cells. The plates were then incubated in a 37 °C incubator with 5% CO2 for 4 hours.

Generation of drug-tolerant persister cells
HCC827 (RRID:CVCL_2063), HCC4006 (RRID:CVCL_1269), and H1975 (RRID:CVCL_1511) were treated with 100 nM, 200 nM, and 500 nM osimertinib, respectively, for 10 days, as described previously(32). The media was replenished every 2 days. Cells were then collected for flow cytometry analysis or for killing assays.

Animal studies
All animal experiments have been approved by the MD Anderson Cancer Center Institutional Animal Care and Use Committee (IACUC). 6- to 8-week-old female NSG (RRID:IMSR_JAX:005557) mice were purchased from Jackson Laboratories. HCC827 cells (3 ×106) or H1975 OR17 cells (3 × 106) were injected subcutaneously into female NSG mice to generate xenograft models as described previously(32). Tumors were measured 3 times per week and were randomized into treatment groups when tumors reached a volume of 60–100 mm3 for the HCC827 model and 75–125 mm3 for the H1975 OR17 model. For CAR-NK cell treatments in HCC827 models, 1 × 107 of control NK cells (non-transduced), CTB-M228BB3z CAR-NK cells, or PNB-M228BB3z CAR-NK cells were i.v. injected in 100 μL HBSS each week for 3 weeks. For CAR-NK cell treatments in H1975 OR17 models, 1 × 107 of control NK cells (non-transduced), PNB-M228BB3z CAR-NK cells, or DNR/PNB-M228BB3z CAR-NK cells were i.v. injected in 100 μL HBSS each week for 3 weeks. 10,000 units of IL-2 and 100 ng of IL-15 were i.p. injected in 200 μL HBSS on the same day and the next day of NK cells treatment. For osimertinib treatment in H1975 OR17 models, mice were dosed with 5 mg/kg osimertinib by oral gavage five days per week for 4 weeks. Body weight and tumor volumes were measured 3 times per week. Mice were humanely euthanized if tumor sizes exceeded 2,000 mm3.
To establish DTP models, H1975 (3 × 106) was subcutaneously injected into female NSG mice (RRID:IMSR_JAX:005557) initially. Tumors were measuredtwice a week and were randomized into treatment groups when they reached a volume of 600 mm3. For osimertinib treatment in the indicated groups, mice were dosed with 5 mg/kg osimertinib by oral gavage five days per week for 2 weeks. For CAR-NK cell treatments, 1 × 107 of control NK cells (non-transduced) or DNR/CTB-M228BB3z CAR-NK cells were i.v. injected in 100 uL HBSS each week for two times at the indicated time. Body weight and tumor volumes were measured 3 times per week. Mice were humanely euthanized if tumor sizes exceeded 2,000 mm3 or ulceration exceeded 5 mm.

IHC staining of CD56
For IHC analysis of CD56, formalin-fixed paraffin-embedded slides were prepared, deparaffinized, and rehydrated as previously described(32). Antigen retrieval was performed using Tris-EDTA Buffer pH 9.0 (ab93684, Abcam) in a steamer for 35 minutes, followed by cooling to room temperature. Endogenous peroxidases were blocked using Endogenous Blocking Solution (SP-6000–100, Vector Laboratories). Blocking was performed using a solution of 5% normal horse serum and 1% normal goat serum in PBS overnight at 4 °C. For CD56 staining in lung cancer patient tissues, anti-CD56 antibodies (clone MRQ-42, RRID:AB_2941091) were used at a dilution of 1:100 in blocking buffer at room temperature for 30 minutes. After washing in PBS, ImmPRESS horse anti-rabbit IgG Polymer Reagent (MP-7401; Vector Laboratories) was applied according to the manufacturer’s instructions. After washing in PBS, the slides were developed using an ImmPACT DAB substrate kit (Cat#SK-4105, Vector Laboratories) for 5 minutes.

Bioinformatic analysis of public data sets
Gene expression data sets GSE123031, GSE125365, GSE121634, GSE146850, GSE210549, GSE201608, and GSE202859 were downloaded from the GEO repository (https://www.ncbi.nlm.nih.gov/geo). Our gene expression data set GSE121634 was reused. All data analysis was performed using R (v.4.2). EMT scores were calculated based on a 76-gene EMT signature(7). Microarray and RNA-seq data were re-analyzed by comparing resistant cells or DTPCs with the parental cells using limma (RRID:SCR_010943) package(33). Heatmaps were generated using logFoldChange (logFC) of indicated genes by the ComplexHeatmap (RRID:SCR_017270) package(34).

Statistics
Data were represented as means ± SD. The statistical significance was determined using Student’s t test (two-tailed), one-way ANOVA, and two-way ANOVA, as appropriate. All statistical analyses were performed by GraphPad Prism 9 (RRID:SCR_002798). Area under the curve (AUC) was calculated by GraphPad Prism 9 (RRID:SCR_002798). p ≤ 0.05 is considered statistically significant, unless otherwise specified.

Data and materials availability
All data associated with this study are present in the paper or the Supplementary Materials. All reagents and materials supporting the findings of this study are available and will be freely distributed by the corresponding authors upon request.

RESULTS

EGFR CAR-NK and CAR-T cells exert potent killing activity against EGFR positive NSCLC cells
We generated EGFR-targeting CARs using sequences of single-chain variable fragment (scFv) derived from two distinct anti-EGFR antibodies, cetuximab (CTB) and panitumumab (PNB). We composed the scFvs into a second-generation CAR construct, CAR-8BB3z, and two third-generation CAR constructs, CAR-828BB3z and CAR-M228BB3z, respectively (Fig. S1A). Next, to generate EGFR CAR-T cells, we activated human T cells from peripheral blood mononuclear cells (PBMCs) using anti-CD3/CD28 Dynabeads and then transduced the T cells with CAR-8BB3z, CAR-828BB3z, and CAR-M228BB3z retrovirus. Despite the absence of prior T cell purification, the expanded cell populations consistently achieved >99% purity of T cells, as confirmed by CD3 positive on flow cytometry (Fig. S1B). There was a higher proportion of CD4+ population in the second-generation 8BB3z CAR-T cells, whereas both third-generation CARs were enriched for CD8+ T cells (Fig. S1C). We utilized anti-mouse IgG F(ab’)2 antibodies for CTB scFv and anti-human IgG F(ab’)2 antibodies for PNB scFv by flow cytometry analysis to assess the expression of CAR molecules, with 54.2% to 79.8% of T cells being CAR positive and the rest lacking the CAR expression (Fig. S1D). The transduction efficiency was similar among different CAR generations and between CTB- and PNB-derived CARs. However, the second-generation 8BB3z CARs had higher MFI compared to the third-generation CARs, while the M228BB3z CAR had slightly higher MFI compared to 828BB3z CARs (Fig. S1D). Furthermore, we used fluorescent-labeled recombinant EGFR protein to assess the binding ability of CAR-T cells and found that all CAR-positive cells showed strong EGFR binding capacity as detected by flow cytometry (Fig. S1E). Moreover, the EGFR binding capacity of different CAR generations correlated with the expression level of CAR on the CAR-T cell surface.
In parallel, we developed a CAR-NK cell transduction and expansion system to generate EGFR CAR-NK cells. K-562 cell-based feeder cells were established by over-expressing 4–1BBL, OX40L, CD86, and membrane-bounded IL-21(35–37) (Fig. S2A&B). After activation with the feeder cells for 4 days, NK cells were retrovirally transduced with CTB- and PNB-derived CARs using a similar approach for generating CAR-T cells (Fig. S2C). The expanded NK cells achieved about 75% purity of NK cells on day 7 and more than 95% on day 14, as confirmed by CD3- CD56+ on flow cytometry (Fig. S1B). Expression of the CAR on NK cells was assessed by flow cytometry following staining with anti-human/-mouse IgG F(ab’)2 antibodies as described above (Fig. S2E). We observed a similar transduction efficiency among different CAR generations, with 67.1% to 74.8% of CTB-derived CARs and 79.3% to 81.4% of PNB-derived CARs. The second-generation 8BB3z CARs had higher MFI compared to the third-generation CARs, while the M228BB3z CAR had slightly higher MFI compared to the 828BB3z CARs (Fig. S2E).
Next, we wanted to make sure the EGFR CAR-T and CAR-NK cells killed EGFR-expressing cell lines before investigating their activity against DRCs and DTPCs. We assessed the cytotoxicity of these EGFR CAR-T cells against a panel of NSCLC cell lines with differential levels of EGFR expression (Fig. S3). Tumor cells were transduced to stably express firefly luciferase, and cytotoxicity assays were performed by co-culturing luciferase-expressing tumor cells with CAR-T cells. CTB- and PNB-derived EGFR CAR-T cells killed both HCC827 (EGFR high) and H1299 (EGFR moderate) cell lines (Fig. 1A&B), with greater killing was observed with HCC827 cells.
Next, we assessed the cytotoxic activity of EGFR CAR-NK cells against HCC827 (EGFR high), H1975 (EGFR high), and HCC4006 cells (EGFR moderate) (Fig. S3). CTB-M228BB3z CAR-NK cells displayed enhanced killing activity compared to CTB-828BB3z or CTB-8BB3z CAR-NK cells across all three EGFR positive cell lines (Fig. 1C). Likewise, PNB-M228BB3z CAR-NK cells exhibited increased killing activity compared to other PNB-derived CAR-NK cells (Fig. 1D). We then assessed the cytotoxic activity of EGFR CAR-NK cell killing activity against additional NSCLC cell lines, including A549, H322, H1299, and Calu-3, expressing differential levels of EGFR (Fig. S3) and observed that some NSCLC cell lines, including H322 and H1299, were highly sensitive to unmodified NK cell-mediated killing (Fig. 1E), suggesting that a subset of NSCLC tumors may be particularly sensitive to NK-based targeting approaches. Taken together, these data demonstrate that EGFR CAR-T and CAR-NK cells have significant antitumor activity against NSCLC cells expressing EGFR, with the third-generation M228BB3z CAR-NK cells having the highest activity among the constructs tested. It also indicates that CAR-independent activation pathways in NK cells contribute to the cytotoxicity against NSCLC cells.

EGFR CAR-NK cells exhibit enhanced killing activity against NSCLC with acquired resistance to osimertinib
We next investigated the activity of EGFR-directed cellular therapies for EGFR mutant NSCLC cells resistant to osimertinib, a standard first-line EGFR TKI. We tested EGFR CAR-T and CAR-NK cells using a panel of parental and osimertinib-resistant (OR) cells we previously developed including H1975 OR5, H1975 OR16, HCC4006 OR2, and HCC4006 OR7 cells all of which developed EMT-mediated EGFR TKI resistance (11,32). 8BB3z EGFR CAR-T cells showed specific lysis activity against H1975 OR5 and OR16 cells, although the killing activity was decreased as compared to H1975 parental cells (p < 0.001 for CTB-8BB3z and p = 0.005 for PNB-8BB3z; Fig. 2A). Likewise, third-generation CTB-M228BB3z and PNB-M228BB3z CAR-T cells demonstrated specific lysis of HCC4006 OR cells, but the OR cells exhibited reduced cell lysis as compared to HCC4006 parental cells (p < 0.001) (Fig. 2B). These data suggested that NSCLC cells with acquired EGFR-TKI resistance may be less sensitive to EGFR CAR-T cells. Using gene expression data from our previous RNA-seq datasets, we compared the transcriptional profiles of EGFR TKI-resistant cells with those of parental cells and found that the EGFR RNA level was decreased in NSCLC cell lines with acquired resistance to erlotinib (Fig. S4A&B). As the activity of CAR-T cells is highly dependent on target antigen expression(38,39), we next assessed cell surface EGFR expression level on parental and OR cells by flow cytometry. H1975 OR cells showed lower levels of cell surface EGFR compared to parental H1975 cells (Fig. S4C). However, HCC4006 OR cells displayed similar levels of EGFR as parental HCC4006 cells (Fig. S4C).
Next, we performed long-term killing assays by co-culturing third-generation EGFR CAR-T cells with parental H1975 and H1975 OR5 cells. After 48 hours, CAR-T cells were harvested and evaluated by flow cytometry for expression of exhaustion markers. We observed a significant upregulation of PD-1 on both CD4+ and CD8+ T cells exposed to H1975 OR5 cells as compared to those co-cultured with parental H1975 cells (Fig. 2C and S4D). In addition to PD-1, LAG-3 and TIM3 were significantly upregulated on CD4+ T cells after co-cultured with H1975 OR5 cells. TIGIT expression was slightly reduced on both CD4+ and CD8+ T cells co-cultured with H1975 OR5 cells (Fig. S4E&F).
Second-generation EGFR CAR-NK cells exhibited significantly increased specific cell lysis against H1975 OR cells and HCC4006 OR cells compared to their respective parental cells (Fig. 2D). Similarly, we observed increased killing activity of third-generation EGFR CAR-NK cells against OR cells compared to parental lines (Fig. 2E-H). Notably, H1975 OR cells and HCC4006 OR cells displayed enhanced sensitivity to non-transduced control NK cells (Fig. 2E-H), suggesting that while EGFR TKI resistant cells may exhibit decreased vulnerability to T cells, they may exhibit heightened sensitivity to NK cell targeting.

Augmented natural cytotoxicity of NK cells contributes to enhanced killing of CAR-NK cells against acquired EGFR TKI-resistant NSCLC
Previous studies have indicated that NK cells can exert their cancer cell elimination capabilities through both CAR-dependent and CAR-independent mechanisms(26,40). With this understanding, we aimed to explore whether the natural cytotoxic activity contributes to the enhanced killing exhibited by CAR-NK cells against acquired EGFR TKI-resistant NSCLC. We interrogated expression changes data utilizing microarray gene expression data and RNA sequencing (RNA-seq) data from the Gene Expression Omnibus (GEO) database. Our analysis revealed notable upregulation of NCR3LG1 (B7-H6), a ligand for the natural cytotoxicity triggering receptor-3 (NCR3/NKp30), along with NKG2D ligands including MICA, MICB, and ULBP1, in NSCLC cells with acquired resistance to EGFR TKIs (Fig. 3A). Notably, the expression of NCR3LG1 has been linked to a mesenchymal gene expression signature(41). Our analysis using the Cancer Cell Line Encyclopedia (CCLE) dataset showed a positive correlation between NCR3LG1 expression and EMT scores in NSCLC cells (R = 0.38, p < 0.001; Fig. S5A). Moreover, we observed that EGFR-TKI-resistant cells, which had undergone EMT, exhibited a trend towards increased expression of NCR3LG1 (Fig. 3A). Gene expression microarray analysis further indicated that MICA, MICB, ULBP1, and/or ULBP2 were upregulated following induction of EMT by TGF-β1 treatment, overexpression of the EMT transcription factor SNAI1, or knockdown of CDH1 in HCC827 cells (42) (Fig. S5B). Similarly, while treatment with TGF-β3 for 4500 hours resulted in EMT in H358 cells, we observed the upregulation of MICA, MICB, ULBP1, ULBP2, ULBP3, and NCR3LG1, along with the increased EMT scores (Fig. S5C). These findings suggest that the shift towards a mesenchymal phenotype is associated with increased expression of NK activating ligands, consistent with our prior observations in the broader NSCLC population(43).
Next, we assessed the expression of B7-H6, MICA/B, and ULBP1 using flow cytometry and observed marked increases in their expression on H1975 OR5 and OR16 cells compared to H1975 parental cells (Fig. 3B&C). Similar significant upregulation was observed for MICA/B on HCC4006 OR2 and OR7 cells, while B7-H6 exhibited upregulation in HCC4006 OR2 cells, and ULBP1 was upregulated in HCC4006 OR7 cells (Fig. S5D).
To assess the impact of these ligands on the sensitivity of EGFR TKI-resistant cells to NK and CAR-NK cells, we utilized blocking antibodies for the corresponding receptors, NKp30 and NKG2D, on control NK (NT-NK) and EGFR-CAR NK cells. Blockade of NKp30 or NKG2D led to a significant reduction in the killing activity of both control NK cells and EGFR-CAR NK cells against EGFR TKI-resistant cells (Fig. 3D). While the double blockade of NKp30 and NKG2D abolished the cytotoxicity of control NK cells, for EGFR-CAR NK cells dual blockade of NKp30 and NKG2D retained specific killing activity against parental and OR cells (Fig. 3D). Moreover, in the context of HCC4006 cells, the blockade of NKG2D led to a reduction in the cytotoxicity of EGFR-CAR NK cells against both parental and OR cells, while also completely inhibiting the activity of control NK cells in similar conditions. In contrast, the impact of NKp30 blockade on the activity of NK or CAR-NK cells against HCC4006 parental or resistant cells was relatively minor (Fig. 3E).
NKG2A is an inhibitory receptor expressed on NK cells and blocking this signaling has been reported to enhance NK cell-mediated tumor immunity(44). To investigate the impact of NKG2A on EGFR-CAR NK cells within the context of EGFR TKI resistance, we employed the anti-NKG2A antibody monalizumab in our co-culture killing assays. Monalizumab did not exhibit significant effects on either control NK cells or EGFR-CAR NK cells when co-cultured with either osimertinib-naïve cells (H1975 and HCC4006) or osimertinib-resistant cells in vitro (Fig. S5E&F).
In summary, the observed upregulation of ligands for NKp30 and NKG2D on EGFR TKI-resistant cells contribute to their increased sensitivity to NK and CAR-NK cells. These findings demonstrate that EGFR CAR-T and CAR-NK cells are effective in targeting EGFR TKI-resistant NSCLC cells, and importantly, these resistant cells exhibit an increased responsiveness to CAR-NK cells.

EGFR TKI DTPCs are sensitive to EGFR CAR-T and CAR-NK cells in vitro
EGFR TKI DTPCs are a subpopulation of cancer cells that survive initial treatment with EGFR TKIs and subsequently enter a quiescent state, persisting for an extended period of time. These DTPCs eventually resume exponential growth and give rise to drug resistant cells. We next assessed whether EGFR CAR-based approaches could target EGFR TKI DTPCs. We treated HCC827, HCC4006, and H1975 with osimertinib for 10 days at which point only residual DTPCs remained. Using a 4-hour luciferase reporter killing assay, we observed that both EGFR CAR-T (Fig. 4A) and CAR-NK cells (Fig. 4B) have potent activity against EGFR TKI DTPCs. Notably, these EGFR CAR-engineered cells exhibited an enhanced killing activity against DTPCs as compared to treatment naïve parental cells.
Next, we observed that EGFR expression was significantly elevated on the cell surface of DTPCs as compared to parental cells using flow cytometric analysis (Fig. 4C). In addition, osimertinib-induced DTPCs showed increased B7-H6 expression (Fig. S6A and 4D), as well as increased expression of ULBP1 and MICA/B (Fig. 4E&F and S6B&C). Next, we utilized anti-NKp30 and anti-NKG2D blocking antibodies in combination with NK cells and observed that the cytotoxicity of control NK cells against HCC4006 parental cells and DTPCs was completely abrogated (Fig. 4G). EGFR-CAR NK cells subjected to this double blockade retained robust killing activity against HCC4006 DTPCs, although reduced level was observed compared to EGFR-CAR NK cells without blockade (Fig. 4G). Similar results were observed when H1975 parental and DTPCs were used as target cells (Fig. 4H). These data indicate that EGFR TKI DTPCs can be targeted by EGFR-CAR NK cells, involving both CAR-dependent and natural cytotoxic receptor-dependent mechanisms.

Osimertinib potentiates EGFR CAR-T and CAR-NK cell mediated killing activity against EGFR TKI resistant cells
Previous studies by our group and others have shown that treatment with HER2 TKIs resulted in accumulation of HER2 on the tumor cell surface, which enhances binding and anti-tumor effects of anti-HER2 antibody-based therapies (45–47). Elevation of EGFR on the cell surface of osimertinib-derived DTPCs suggests that osimertinib may similarly induce EGFR accumulation. Therefore, we next hypothesized that EGFR TKIs such as osimertinib could enhance the efficacy of EGFR CAR-based cell therapy for EGFR mutant NSCLC tumor cells with acquired EGFR TKI resistance. We cultured H1975 OR and HCC4006 OR cells without osimertinib for 3 days, then treated H1975 OR and HCC4006 OR cells with osimertinib for 24 hours and assessed EGFR expression by flow cytometry. In all cell lines tested, we observed upregulation of EGFR on the cell surface following treatment with osimertinib (Fig. S7A). As expected, osimertinib did not impact the survival of OR cells (Fig. S7B). Cytotoxicity assays revealed that the killing activity of EGFR CAR-T cells was higher in OR cells treated with osimertinib as compared to untreated cells (Fig. S7C). Likewise, EGFR CAR-NK cells showed increased activity against H1975 OR and HCC4006 OR cells treated with osimertinib (Fig. S7D). Taken together, these findings indicate that the combination of osimertinib and EGFR CAR-T or CAR-NK cells can enhance the killing of EGFR mutant NSCLC cells, compared with CAR-T or CAR-NK cells alone, despite the resistance of these cells to osimertinib. This finding supports combination regimens in which the TKI is continued during EGFR CAR-NK or CAR-T-based therapies.

Blockade of TGF-β signaling enhanced the activity of EGFR CAR-NK cells against EGFR TKI resistant cells
The immunosuppressive TME is a major challenge for CAR-based cell therapy approaches to treat solid tumors. TGF-β is a potent immunosuppressive cytokine secreted by tumor cells that attenuates the immune response by suppressing T cell activation and proliferation and inhibiting CD8 T and NK cell cytotoxic function. We have previously demonstrated the upregulation of TGF-β members (TGFB1 and TGFB2) and their receptor (TGFBR3) in EGFR-mutant tumors compared to EGFR wild-type tumors(48), which indicates its potential role in modulating responses of CAR-based therapies. To investigate the role of TGF-β in EGFR TKI resistance, we performed transcriptomic analysis on both our own and publicly available datasets. Our analysis revealed upregulation of genes in TGF-β family, including TGFB1, TGFB2, and TGFB3 (Fig. 5A), in EMT-associated EGFR TKI resistant cells, consistent with our previous finding(48). In addition, receptors genes TGFBR1 and TGFBR2, along with TGF- β signal transducers of the Smad family (SMAD2, SMAD3, and SMAD4), were also upregulated in EGFR TKI resistant cells compared to their respective parental lines (Fig. 5A). Furthermore, ELISA assay revealed significantly increased secretion of TGF-β1 in the majority of EGFR TKI resistant cells as compared to parental cells (Fig. 5B). Because increased TGF-β signaling may impair the killing activity of CAR-NK cells in NSCLC with acquired TKI-resistance, we assessed whether blockade of TGF-β signaling could enhance CAR-based cell killing. Pharmacologic blockade of TGF-β signaling with galunisertib significantly enhanced killing of H1975 parental and OR cells after 72-hour co-culture with WT NK or EGFR CAR-NK cells (Fig. 5C). TGF-β blockade also impacted the survival of H1975 parental and OR cells, even without effector cells. In contrast, TGF-β blockade enhanced the survival of HCC4006 parental cells but inhibited HCC4006 OR cell survival. TGF-β blockade significantly reduced the survival of HCC4006 OR cells but not HCC4006 parental cells co-cultured with WT NK or EGFR CAR-NK cells (Fig. 5D). These data indicate that blockade of the TGF-β pathway may enhance the activity of EGFR CAR-NK cells against EGFR TKI resistant cells.

Dominant-negative TGF-β receptor CAR-NK cells are resistant to TGF-β-mediated immunosuppression
We next sought to determine whether CAR-NK activity against EGFR mutant NSCLC cells could be enhanced by depleting TGF-β signaling in CAR-NK cells. Thus, we linked the sequence of dominant-negative TGF-β receptor II (DNR) to our EGFR CAR constructs using a self-cleaving T2A peptide. The DNR acts as a trap that binds to TGF-β but does not transduce downstream signaling (Fig. 6A). This DNR is capable of blocking signaling by all three TGF-β isoforms. After retroviral transduction of NK cells, co-expression of the DNR with the CAR was confirmed by flow cytometry using anti-TGF-β receptor 2 antibody on the CAR-NK cells (Fig. 6B). SMAD2/3 are essential intracellular signaling components of the TGF-β pathway. Comparable levels of phosphorylated SMAD2 (p-SMAD2) were detected in NK (CAR negative population) and CAR-NK cells after stimulation with TGF-β1. In contrast, p-SMAD2 levels were decreased in DNR-CAR-NK compared to the CAR-negative population following TGF-β1 stimulation (Fig. 6C). Moreover, TGF-β1 treatment induced significant down-regulation of perforin and granzyme B (GZMB) in WT NK (perforin, p < 0.05; GZMB, p < 0.05) and CAR-NK cells (perforin, p < 0.001; GZMB, p < 0.001), while TGF-β1-treated DNR-CAR-NK cells still had similar levels of perforin as compared to non-treated CAR-NK or DNR-CAR-NK cells (Fig. 6D&E).
We next investigated the impact of TGF-β1 and the DNR on CAR-NK-mediated killing of OR cells. To do this, we treated OR cells with or without exogenous TGF-β1 in the presence of EGFR CAR-NK with or without the DNR. EGFR CAR-NK cells showed decreased killing activity against osimertinib-resistant cell lines with TGF-β1treatment. Notably, DNR-CAR-NK cells showed enhanced killing activity as compared to CAR-NK cells. Moreover, DNR-CAR-NK cells showed similar cytotoxicity with the addition of exogenous TGF-β1 compared to those without TGF-β1 treatment (Fig. 6F). Given our earlier finding that osimertinib could enhance the tumor cell killing of EGFR CAR-based approaches, we tested the combination of DNR-CAR-NK cells with osimertinib. Consistent with our previous results, adding osimertinib enhanced the sensitivity of OR cells to DNR-CAR-NK cells (Fig. S8A&B).

EGFR-CAR NK cells demonstrate potent in vivo activity against NSCLC xenografts with acquired osimertinib resistance
To assess the antitumor activity of EGFR CAR-NK cells in vivo, we initially utilized a human EGFR mutant NSCLC xenograft mouse model by subcutaneously injecting HCC827 cells into the flank of NSG mice. Once the tumors reached approximately 80 mm3, the animals were randomized to receive control human NK cells, CTB-M228BB3z CAR-NK cells, or PNB-M228BB3z CAR-NK cells, with 8 to 10 mice per group. Although tumor growth was also inhibited by WT NK cells in this model (WT NK vs. Control, p < 0.001), EGFR CAR-NK cells resulted in significantly greater regression of tumor volume (CTB-CAR-NK vs. WT NK, p < 0.001; PNB-CAR-NK vs. WT NK, p < 0.001), and tumors were no longer detected by day 50 in all CTB-CAR-NK treated (10/10) and PNB-CAR-NK treated (8/8) mice (Fig. 7A). Next, we evaluated EGFR CAR NK cells in an osimertinib-resistant xenograft model. H1975 OR17 cells were injected subcutaneously into NSG mice, and when tumors reached approximately 95 mm3, animals were randomized to receive vehicle, osimertinib, control NK cells, control NK cells + osimertinib, EGFR CAR-NK cells, EGFR CAR-NK cells + osimertinib, DNR-CAR NK cells, and DNR-CAR NK cells + osimertinib. EGFR CAR-NK cells significantly inhibited tumor growth as compared to control NK cells (p < 0.001). In parallel, adding osimertinib significantly enhanced the cytotoxicity of WT NK cells against osimertinib-resistant cells (p < 0.001). Importantly, DNR-CAR-NK cells demonstrated even greater antitumor efficacy than EGFR CAR-NK cells, with significantly reduced tumor volumes (p < 0.001; Fig. 7B). Collectively, these data indicate that EGFR CAR-NK cells are an effective approach for treating osimertinib-resistant NSCLC in vivo, and the efficacy of this CAR-based approach can be enhanced with the addition of osimertinib and co-expression of DNR in CAR-NK cells.

EGFR-CAR NK cells demonstrate potent in vivo activity against osimertinib DTPCs in xenograft models of NSCLC
To evaluate the antitumor activity of EGFR CAR-NK cells against DTPCs in vivo, we utilized a H1975 DTPC mouse model. Initially, H1975 xenograft tumors were established by subcutaneously injecting H1975 cells into the flank of NSG mice. Once tumors reached approximately 600 mm3, animals were treated with osimertinib for 2 weeks, at which point tumor volumes were an average of 210 mm3. The animals were randomized to receive vehicle, osimertinib, osimertinib + WT NK cells, osimertinib + EGFR CAR-NK cells, osimertinib followed by WT NK cells, and osimertinib followed by EGFR CAR-NK cells. As expected, we observed tumor regrowth upon discontinuing osimertinib treatment (Fig. 7C). Mice concurrently treated with osimertinib and EGFR CAR-NK cells (osimertinib + EGFR CAR-NK cells) exhibited further reductions in tumor volumes, which were an average of 73.2 mm3, and prolonged response durations compared to the osimertinib-treated group (Fig. 7C). Moreover, mice initially treated with osimertinib (DTPCs), and then EGFR CAR-NK cells (osimertinib then EGFR CAR-NK cells) also exhibited further reductions in tumor volumes, which were an average of 66.8 mm3 (Fig. 7D). Furthermore, treatment with EGFR CAR-NK cells significantly prolonged survival time in both groups (Fig. 7E). We analyzed tumors for NK cell infiltration by IHC using antibodies specific for human CD56. Tumors were collected at the endpoint of the experiment, which was 22 days after NK infusion for OSI + NK/CAR-NK groups and 8 days after infusion for OSI then NK/CAR-NK groups. Positive CD56 staining was observed in tumors from mice treated with NK or CAR-NK cells (Fig. 7F). Collectively, these data indicate that EGFR CAR-NK cells are an effective approach for targeting osimertinib-induced DTP of NSCLC in vivo.

DISCUSSION
In NSCLC patients bearing EGFR mutations, there is a major unmet need for effective therapies to target residual disease after initial EGFR TKI treatment as well as EGFR-TKI refractory tumor cells that eventually emerge. In this study, we demonstrated that EGFR CAR-NK and CAR T cell therapies are effective approaches to eliminate EGFR mutant DTPCs and DRCs after EGFR TKI therapy. Our findings reveal that these DTPCs and DRCs demonstrate a shift in their immune phenotype, resulting in a reduced responsiveness to EGFR CAR-T cells but a heightened sensitivity to CAR-NK cells. Furthermore, we report that the cytotoxicity of EGFR CAR-NK cells can be enhanced by combination with osimertinib and the inhibition of the TGF-β singling pathway in CAR-NK cells to overcome the immunosuppressive tumor microenvironment. In addition, we show that EGFR CAR-T and CAR-NK cells can effectively target EGFR TKI DTPCs and thus could potentially be used to eliminate residual disease and prevent the emergence of fully drug resistant tumors. Collectively, our findings indicate EGFR-targeting cellular immunotherapies, particularly CAR-NK cells, may be an effective treatment strategy for EGFR mutant NSCLC patients and that TKI resistance is associated with shifts in immunomodulatory pathways that may promote relative resistance to CAR-T therapies.
Although we observed some non-EGFR mutant cell lines respond to CAR-NK cells in vitro, this study focused specifically on EGFR mutation-positive NSCLC for several reasons. First, the EGFR protein is critical for the survival of EGFR mutant tumors, it is expected that antigen loss– a common potential mechanism of immune escape–is less likely compared to EGFR wild-type tumors. Second, EGFR mutant NSCLC patients are a well-defined subgroup that usually have an initial major response to TKI therapy, and the amount of residual tumor typically at least an order of magnitude lower than the pre-treatment baseline. We hypothesize that targeting residual tumor at this stage of low tumor bulk is more likely to be successful than higher tumor volume disease that is refractory to initial treatment. Finally, there are currently no therapies with proven efficacy against residual disease in EGFR TKI treated patients, and this represents a major unmet need in this population for the field.
Our primary focus was on assessing cytotoxicity and tumor burden as key functional outcomes, especially regarding CAR-NK cells. We measured cytotoxicity using luciferase-based viability assays, which offer a high-throughput, sensitive, and quantitative method for evaluating tumor cell killing. In later stages of the study, we prioritized CAR-NK cells due to their promising therapeutic potential and lower risk of cytokine release syndrome compared to CAR-T cells. Consequently, cytokine secretion profiling was not extensively performed, apoptosis-specific assays were not conducted, and only in vivo studies with CAR-NK cells were carried out. As a result, the comparative therapeutic efficacy and toxicity of CAR-T versus CAR-NK cells were not fully evaluated. Future research involving side-by-side comparisons, cytokine profiling, and apoptosis markers will be essential for a more comprehensive understanding of the mechanisms and safety profiles of each cell therapy approach.
The CAR constructs used in this study, which contain canonical T cell signaling and co-stimulatory domains (CD3ζ, 4–1BB, and CD28), were originally designed for T cells. These constructs were directly adapted for NK cells to evaluate the efficacy of CAR-NK cells against EGFR-expressing tumor cells. While no significant differences in cytotoxicity were observed among the CAR constructs in T cells, CAR-NK cells expressing the M228BB3z construct exhibited the most robust killing of EGFR-positive tumor cells compared to the 8BB3z and 828BB3z constructs. Based on this result, we selected M228BB3z for subsequent experiments, although we didn’t further optimize it for NK cell biology. This finding highlights that CAR design elements can differentially impact efficacy in NK cells. Notably, several clinical studies have already employed T cell-oriented CAR designs in CAR-NK cells, such as CD19-targeted CAR-NK cells using CD28 and CD3ζ domains, demonstrating clinical safety and efficacy in hematologic malignancies(27,49). Unlike T cells, NK cells have distinct mechanisms for activation and signaling requirements. Growing evidence suggests that CAR constructs optimized for NK cell biology, incorporating domains such as NKG2D, 2B4, or DAP10, can significantly enhance CAR-NK cell efficacy(50–53). For example, a CAR construct containing the transmembrane domain of NKG2D, the 2B4 co-stimulatory domain, and the CD3ζ signaling domain exhibited superior antitumor activity(54). Our current study represents a proof-of-concept evaluation of EGFR-directed CAR-NK cells using conventional CAR designs. Moving forward, future work will include engineering NK-tailored CARs to enhance therapeutic efficacy and translational potential.
Our data revealed that EGFR-CAR T cells exhibit robust activity against NSCLC cells with acquired resistance to EGFR-TKIs, although the activity was decreased compared to their performance against TKI-naïve cells. One contributing factor to this reduced potency may be the diminished cell surface expression of EGFR on resistant cells. Such a reduction in EGFR levels impacts the effectiveness of CAR-T cells, which heavily rely on the presence of target antigens for their functionality(38). In addition, EMT appears to be another contributing factor leading to the decreased sensitivity of EGFR TKI-resistant cells to EGFR CAR-T cells. The acquisition of EGFR TKI resistance often involves EMT, a phenomenon characterized by increased motility, invasiveness, drug resistance, and properties similar to those of cancer stem cells(7,8,11,32,55–58). Importantly, EMT has been implicated in promoting an immunosuppressive EMT that limits the activation and function of immune cells. A mesenchymal phenotype is associated with elevated expression of immunosuppressive cytokines (e.g., TGF-β, IL-10, and IL-6) and immune checkpoint molecules (e.g., PD-L1, TIM-3, and CTLA-4) (43,59–61). Consistent with these mechanisms, flow cytometric analysis in our study revealed upregulation of PD-1 on both CD4+ and CD8+ CAR-T cells co-cultured with EGFR TKI-resistant tumor cells, compared to those exposed to drug-sensitive cells. Furthermore, mesenchymal-like transcriptional signatures have been shown to be associated with ICI resistance in multiple clinical studies (41). These observations align with the clinical findings of reduced efficacy of ICI in EGFR TKI-resistant NSCLC cases(16–21).
In contrast, our data indicate that EGFR-CAR NK cells exhibited increased activity against EGFR TKI-resistant cells. We demonstrate that both CAR-dependent mechanisms and CAR-independent natural cytotoxicity pathways contribute to the killing activity of EGFR-CAR NK cells. Our data also indicates that EMT contributes to the augmented sensitivity of EGFR TKI-resistant cells towards EGFR CAR-NK cells. Notably, cells with mesenchymal transcriptional signatures exhibit greater susceptibility to NK cell-mediated killing compared to cells with an epithelial transcriptional signature (41). This sensitivity discrepancy likely involves multiple mechanisms, including an elevated expression of B7-H6 (NCR3LG1), a ligand that triggers the NKp30-mediated activation of NK cells (62,63). Notably, cells that have undergone EMT due to acquired EGFR TKI resistance, such as H1975 OR and HCC4006 OR cells (11,32), exhibit upregulation of B7-H6.
In addition to B7-H6 (NCR3LG1), the increased expression of NKG2D ligands such as MICA, MICB, and ULBP1 also contributes to the increased sensitivity of EGFR TKI-resistant cells to NK cells. These NKG2D ligands, which are typically absent in most normal tissues, are induced in response to cellular stress, including viral infections or malignant transformation. Importantly, our findings underscore the capacity of EGFR TKIs to stimulate the upregulation of these ligands in cells harboring EGFR mutations. Collectively, these findings suggest potential advantages for EGFR CAR-NK cell therapy over CAR-T cells in treating NSCLC with acquired EGFR TKI-resistance, particularly in the context of EMT. Furthermore, our findings emphasize the important role of natural cytotoxicity mediated by CAR-NK cells in anti-tumor activity against DTPCs and DRCs.
Distinct from fully DRCs, DTPCs are thought to be more flexible and transient in their resistant state that can be reversed by discontinuing treatment with the same drug. DTPCs may remain quiescent or clinically invisible for prolonged periods, but eventually progress to become fully resistant cells that resume growth and metastasis(64). In this study, we observed that the DTPCs of osimertinib-treated EGFR mutant NSCLC cells displayed elevated sensitivity to EGFR CAR-T cells, likely due to the upregulation of EGFR on the cell surface following osimertinib treatment, which is different from the fully drug resistant cells. Moreover, DTPCs also showed enhanced sensitivity to EGFR CAR-NK cells, likely partly due to the upregulated expression of B7-H6, ULBP1, and MICA/B. Thus, osimertinib can sensitize DTPCs to both EGFR CAR-T and CAR-NK cells. These findings prove that CAR-based cellular therapies may be effective therapeutic approaches for targeting DTPCs. Moreover, it suggests that the combination of EGFR CAR-T/-NK cell therapies with EGFR-TKIs may effectively delay the emergence of EGFR TKI resistance as compared to TKIs alone.
It is notable that, particularly when using certain cell lines and DTPCs as targets, WT NK cells themselves can exhibit significant innate cytotoxic activity (far greater than control T cells) and therefore they are not a “negative control” but rather just a control to highlight the added activity that comes from targeting EGFR specifically. This activity is largely mediated by NCR3 and NKG2D, due to high levels of their corresponding ligands expressed on these tumor cells. As a result, the use of WT NK cells as a control may lead to CAR-NK cells appearing only marginally more effective. While additional strategies such as antibody blockade of the CAR–EGFR interaction or the use of EGFR-knockout tumor cells could help dissect antigen-specific effects, these approaches may also compromise tumor cell viability and confound interpretation. Using irrelevant targets, such as CD19-targeting CAR, would be another approach as a control although it would be expected to demonstrate the same innate cytotoxic activity as the NK cells we used as well as potential added activity if the “irrelevant target” is expressed even at low levels on tumor cells. Therefore, the WT NK cells used here are a more appropriate control. Furthermore, the fact that innate cytotoxicity of NK cells is augmented by changes induced with osimertinib resistance is a noteworthy point and one that adds to, not detracts, from the findings of the study.
In this study, we also observed that osimertinib induced the upregulation of EGFR on the cell surface of osimertinib-resistant NCSLC cell lines. Previous studies from our group and others have similarly shown that HER2 TKIs cause accumulation of HER2 on the tumor cell surface, which then enhances the binding and anti-tumor effects of anti-HER2 antibody and antibody-drug-conjugate (ADC) therapies (45–47). Osimertinib-induced upregulation of EGFR on the cell surface of treatment naïve tumor cells and DTPCs may be due to the accumulation of EGFR by the covalent binding of osimertinib. In addition, we found that osimertinib induced the upregulation of EGFR on the cell surface of EGFR TKI-resistant cells, although EMT-related osimertinib-resistant cells are not dependent on EGFR signaling for survival. Treatment with osimertinib sensitized osimertinib-resistant cell lines to EGFR CAR-T and CAR-NK cells-mediated killing in vitro. These findings were supported by our in vivo observation that the combination of osimertinib with EGFR CAR-NK enhanced the anti-tumor activity in osimertinib-resistant NSCLC xenografts. These data indicate that the combination of osimertinib with EGFR CAR NK cells may effectively target EGFR DTPCs and TKI-resistant cells in EGFR mutant NSCLC patients.
Our resistant cell lines were generated by treating tumor cells with increasing concentrations of EGFR KTIs over several months, while DTPCs were developed after 10 to 14 days of osimertinib treatment. Although these models are powerful tools for dissecting specific molecular pathways in a controlled setting, they do not completely recapitulate the complex interplay of factors present in patient tumors under intense selective pressures over prolonged treatment periods and within a dynamic tumor microenvironment, including immune system pressure and stromal interactions. Therefore, our findings suggest that EGFR CAR-NK cells remain a viable therapeutic option in DRCs and DTPCs and should be viewed as a foundational step. Future testing of EGFR-CAR NK cells in more clinically relevant models, such as patient-derived xenografts or organoids, will be crucial to confirm their clinical relevance.
Data presented here and in previously published studies indicate that the TGF-β pathway is activated in NSCLC cells that have developed resistance to EGFR TKIs(32). TGF-β is a potent initiator of EMT and promotes an immunosuppressive TME, exerting both direct and indirect inhibitory effects on T cells and NK cells(65–68). We show that TGF-β secretion impedes the efficacy of EGFR CAR-T and CAR-NK cells, as evidenced by our observation that inhibition of the TGF-β signaling pathway using the TGF-β inhibitor galunisertib improved their functionality. In prior studies inhibition of TGF-β signaling, achieved through approaches like over-expressing DNR or eliminating TGFBR2 via CRISPR/Cas9, has been shown to significantly enhance the control of solid tumors by CAR-T cells(69–71). Similarly, engineering NK cells or CAR-NK cells to overexpress DNR has been shown to result in increased efficacy against tumors thriving in TGF-β-rich environments (72–74). In the present study, we demonstrated that DNR co-expression in EGFR-CAR NK cells leads to significant enhancement in their anti-tumor activity against EGFR-TKI-resistant NSCLC both in vitro and in vivo. These findings indicate that targeting the TGF-β pathway to overcome the immunosuppressive TME can significantly enhance the therapeutic effects of EGFR CAR-NK cells for EGFR TKI-resistant NSCLC, and thus, DNR-expressing EGFR CAR-NK cells could be a promising therapeutic strategy.
One major concern that may limit the application of EGFR CARs in the clinic is the possibility of toxicities arising from the reactivity of these CARs against EGFR expressed in normal tissues, including the skin and gastrointestinal tract. Clinical experience with EGFR-targeting monoclonal antibodies such as cetuximab and panitumumab has demonstrated such toxicities, including dermatologic and mucosal side effects(75–77). Therefore, CAR constructs incorporating scFvs derived from these antibodies could potentially exhibit similar adverse effects. However, because the EGFR-specific scFvs used in our CAR constructs do not cross-react with murine EGFR and the lack of human MHC-I in mouse tissues, the NSG xenograft models used in this study are not suitable for assessing on-target/off-tumor toxicity of EGFR CAR-NK cells. Nevertheless, phase I clinical trials of EGFR CAR-T cells have primarily reported tolerable low-grade dermatologic and mucosal toxicities, with grade 3–4 mucocutaneous events occurring in only a minority of patients(78–81). The use of CAR-NK cells may mitigate this concern because MHC class I (MHC-I) expression in normal tissues, which bind to NK inhibitory receptors(82,83), can keep NK cells quiescent. Consistent with this possibility, CD19 CAR-NK cells in the clinic appear to have fewer immune-related adverse events thus far(49,84). While clinical data on EGFR-specific CAR-NK cells are not yet available, trials using EGFR-targeted NK cells engager AFM24 in combination with autologous NK cells (SNK01) have demonstrated excellent safety, with only grade 1–2 events(85). These findings suggest that EGFR-targeting cell therapies may be clinically feasible despite EGFR expression in some normal tissues. If targeting CARs against EGFR does not prove to be feasible, the observations here regarding the immune phenotype of DTPCs and DRCs can help inform CAR approaches targeting alternative cell surface targets.
In summary, we provide preclinical evidence for the potential superiority of CAR-NK cell therapy over EGFR CAR-T cells in treating NSCLC with acquired EGFR TKI-resistance, particularly in the context of EMT. Additionally, our findings highlight the promise of combining EGFR CAR-based therapies with osimertinib or TGF-β blocking strategies to amplify their efficacy against fully TKI-resistant cells and DTPCs. These data support future clinical studies to evaluate the potential of EGFR CAR-NK cells to prevent and combat EGFR TKI resistant disease.

Supplementary Material
12345678