Genetic Engineering of Tumor‐Infiltrating Lymphocytes (TIL) via a T‐Editor Platform to Enhance Anti‐Tumor Activity

ABSTRACT

Tumor‐infiltrating lymphocytes (TIL) therapy has demonstrated clinical potential in malignancies. However, limited understanding of why only a subset of patients respond to TIL therapy, coupled with the lack of simple and efficient methods to genetically engineer fragile TIL, has hindered efforts to enhance TIL efficacy through genetic modification. A T‐Editor platform enabling rapid and efficient CRISPR‐mediated gene editing in TIL was developed and optimized. To minimize the risk of chromosomal translocations associated with Cas9‐induced double‐strand breaks (DSBs), single‐guide RNAs (sgRNAs) were designed for cytosine base editing (CBE). The expansion capacity, phenotypic profile, cytokine production, and in vitro cytolytic activity of base‐edited TIL were compared with those of Cas9‐KO TIL. In vivo efficacy was assessed using patient‐derived xenograft (PDX) mouse models. The T‐Editor platform was optimized for TIL gene editing by refining stimulation conditions, electroporation parameters, and CRISPR/Cas9 reagent dosing. FAM84B emerged as the top candidate, with its knockout resulting in the most pronounced enhancement of TIL cytolytic activity. CBE‐mediated C·G‐to‐T·A conversion in the FAM84B exon achieved high editing efficiency with minimal insertion–deletion (indel) events. Base‐edited TIL exhibited comparable expansion, phenotype, cytokine production, and in vitro cytolytic activity relative to Cas9‐KO TIL. Compared with non‐engineered control TIL, FAM84B‐edited TIL displayed an increased CD62L⁺ memory subset, enhanced effector function and cytolytic activity, and improved in vivo antitumor efficacy. In conclusion, the T‐Editor platform enables rapid and efficient CRISPR‐mediated gene editing for engineering TIL to enhance its therapeutic potency. FAM84B may represent a novel potential target for improving TIL‐mediated antitumor activity.

A rapid and efficient CRISPR‐mediated gene editing platform for TIL engineering identified FAM84B as a novel potential target to enhance antitumor activity and pioneered the use of CBE to generate FAM84B loss‐of‐function TIL with enhanced antitumor activity.

Abbreviations

ACT

adoptive cell therapy

ATCC

American Type Culture Collection

CBE

Cytosine Base Editing

CR

complete responder

CRISPR

clustered regularly interspaced short palindromic repeats

CTLs

cytotoxic T lymphocytes

DEGs

differentially expressed genes

DSBs

double‐strand breaks

ELISA

Enzyme‐Linked Immunosorbent Assay

FDR

false discovery rate

GMP

Good Manufacturing Practice

gRNA

guide RNA

indel

insertion–deletion

KO

knockout

LDH

lactate dehydrogenase

NE

non‐edited

NR

non‐responders

NSCLC

non–small cell lung cancer

ORR

objective response rate

PBMCs

Peripheral blood mononuclear cells

PD

progressive disease

PDX

patient‐derived xenograft

PE150

paired‐end 150 bp

PR

partial responders

pre‐REP

pre‐rapid expansion protocol

R

responders

REP

rapid expansion

RNP

ribonucleoprotein

rRNA

ribosomal RNA

SD

stable disease

sgRNAs

single‐guide RNAs

TCRs

T cell receptors

TIL

tumor‐infiltrating lymphocyte

1. Introduction

Tumor‐infiltrating lymphocyte (TIL) therapy represents a pivotal form of adoptive cell therapy (ACT) for solid tumors. It involves the isolation of tumor‐infiltrating T lymphocytes from a patient's resected tumor or core needle biopsy, ex vivo expansion of these cells, and their re‐infusion into the patient following lymphodepleting chemotherapy, with interleukin‐2 (IL‐2) administered to support the in vivo expansion and persistence of TIL [1, 2, 3, 4]. TIL therapy has demonstrated promising clinical outcomes in certain settings. For instance, an objective response rate (ORR) of approximately 30% has been reported in patients with refractory metastatic melanoma who had progressed on multiple prior therapies, including immune checkpoint inhibitors [5, 6]. In 2024, the U.S. FDA approved the first TIL therapy, lifileucel, for advanced melanoma, confirming that naturally selected, tumor‐reactive T cells can mediate robust anti‐tumor responses. In patients with cervical cancer or non–small cell lung cancer (NSCLC), ORRs of 30%–45% and 21.4%, respectively, have been observed [7, 8, 9].

Multiple factors influence the prognosis of TIL therapy [10, 11], including baseline tumor mutational burden [12], lactate dehydrogenase (LDH) levels [6], sum of diameters of target lesions [6], immune cell interactions within the tumor microenvironment [12], prior treatment history [13, 14], administered IL‐2 dosage [15], and characteristics of the TIL infusion product [16]. Although several TIL surface markers have been associated with therapeutic response [16, 17], it remains unclear which specific pathways or genes represent the most promising targets for enhancing TIL efficacy against solid tumors—limiting the potential for genetic engineering to improve TIL function. Functional screening using the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 gene‐editing system may offer a powerful approach to identify novel regulators that enhance or impair TIL anti‐tumor activity [18]. However, such screens are technically complex, time‐consuming, and have been conducted primarily in vitro or in mouse models using antigen‐specific cytotoxic T lymphocytes (CTLs) or TCR‐engineered T cells, which may not fully recapitulate the biology of TIL therapy in human cancer patients [19].

TIL are typically exhausted, highly differentiated, and exhibit relatively low viability during and after the pre‐rapid expansion protocol (pre‐REP) culture phase, rendering them susceptible to damage during genetic manipulation. Consequently, electroporation or viral transduction methods yield low gene editing or transduction efficiency and are associated with substantial T‐cell loss [20]. Moreover, most published protocols for TIL genetic modification require a 2‐ to 3‐day stimulation with anti‐CD3 and anti‐CD28 antibodies following pre‐REP and prior to genetic engineering, which further exacerbates cell loss [20, 21, 22, 23]. Therefore, a simplified, efficient, and streamlined genetic engineering approach is needed to facilitate both functional gene studies in TIL and the development of next‐generation TIL‐based therapies.

In this study, we employed a novel approach to explore genes associated with the efficacy of tumor‐infiltrating lymphocytes (TIL) in a real‐world clinical setting by comparing the transcriptomes of infused TIL from responders (R) and non‐responders (NR) enrolled in a Phase I clinical trial. More than 10 differentially expressed genes were identified as potential candidate determinants of response. Although this transcriptomic analysis was a hypothesis‐generating method, these genes could be used to assess TIL cell editing efficacy. We then further developed a T‐Editor platform enabling rapid and efficient CRISPR‐mediated gene editing in TIL, which was optimized to assess the functional impact of these potential candidate genes on TIL expansion, cytokine production, and cytolytic activity. LRATD2 (formerly known as FAM84B) emerged as the top candidate, with its knockout (KO) leading to the most pronounced enhancement in TIL cytolytic activity (Figure 1). To minimize the risk of chromosomal translocations associated with Cas9‐induced double‐strand breaks (DSBs), cytosine base editing (CBE) was utilized alongside traditional Cas9 to achieve FAM84B knockout. The expansion capacity, phenotypic profile, cytokine production, and in vitro cytolytic activity of base‐edited TIL were comparable to those of Cas9‐KO TIL. Compared with non‐engineered control TIL, FAM84B base‐edited TIL exhibited a higher proportion of CD62L+ memory T cells, enhanced effector function and cytolytic activity, and improved anti‐tumor efficacy in vivo.

FIGURE 1.

Graphical abstract. A rapid and efficient CRISPR‐mediated gene editing platform for TIL engineering identified FAM84B as a novel potential target to enhance antitumor activity and pioneered the use of CBE to generate FAM84B loss‐of‐function TIL with enhanced antitumor activity.

2. Methods and Materials

2.1. TIL Manufacturing

TIL products used for patient treatment or in vivo and in vitro studies were manufactured by Suzhou Lanma Biotechnology Company following a standard Good Manufacturing Practice (GMP) procedure as previously described [24]. Briefly, freshly resected tumor tissues were placed in sterile RPMI 1640 medium and transported to the GMP facility. Tumors were dissected into fragments of 2–3 mm³ and seeded into culture vessels (5–10 fragments per vessel) in medium supplemented with 6,000 IU/mL IL‐2 and 10% human AB serum. TIL were passaged into new vessels upon reaching a density of ≥ 2 × 10⁶ cells/mL. Medium containing 6,000 IU/mL IL‐2 was replenished every 2–5 days to support TIL growth over 10–14 days under a pre‐rapid expansion protocol (pre‐REP). TIL from pre‐REP cultures were harvested for subsequent rapid expansion (REP) when patients were ready for treatment. For REP, TIL were cultured in medium containing anti‐CD3 antibody (ACROBiosystems), irradiated feeder cells, and 6,000 IU/mL IL‐2 over a 14‐day period. During REP, TIL were split every 2–3 days upon reaching ≥ 2 × 10⁶ cells/mL. Pathogen detection testing was performed on days 0, 7, and 14 of REP. After completion of the 14‐day REP, TIL were harvested, resuspended in infusion medium, and formulated as final products for infusion.

2.2. TIL Gene‐Editing

CRISPR/Cas9 or CRISPR/cytosine base editor (CBE) ribonucleoprotein (RNP) complexes were formed by incubating 40 pmol Cas9/CBE protein (Kactus Biosystems EE109, KD‐0001) with 80 pmol synthetic gRNA targeting the gene of interest (GenScript) at room temperature for 15 min. TIL were collected after the pre‐REP stage, washed twice with PBS, resuspended in MaxCyte Electroporation Buffer (Cytiva) at a concentration of 5 × 10⁶ cells per 20 μL, and mixed with RNP complexes. Electroporation was performed using the ExPERT ATX Electroporator (MaxCyte). Electroporated TIL were allowed to recover for 1 h before initiation of the REP as described above. Engineered TIL were harvested and resuspended in infusion medium for use in in vitro or in vivo patient‐derived xenograft (PDX) mouse model studies. Final conditions selected for T‐Editor were that TILs were not pre‐stimulated before electroporation, 0.4 nmol sgRNA, and that 0.2 nmol Cas9 protein were the final doses used.

2.3. Gene‐Editing Efficiency Analysis

Genomic DNA was extracted from TIL 3–5 days after electroporation. DNA fragments encompassing the edited region and corresponding wild‐type sequences were amplified by conventional PCR and subjected to Sanger sequencing. Sequencing data were analyzed using ICE (https://ice.editco.bio/) or BEAT (https://www.hanlab.cc/beat/) to determine gene editing efficiency.

2.4. Enzyme‐Linked Immunosorbent Assay (ELISA)

Peripheral blood was collected from patients at different time points according to the protocol. Plasma was isolated from peripheral blood by centrifugation and used for cytokine analysis. The levels of multiple cytokines were measured using pre‐coated ELISA kits (IFN‐γ: Biolegend 430,107; TNF‐α: Dakewe 1,117,202) with standard samples according to the manufacturers' instructions. Absorbance was read at 450 nm using a SpectraMax 190 Microplate Reader (Molecular Devices LLC), and concentration values were calculated using SoftMax Pro 7.1.2 (Molecular Devices LLC).

2.5. Flow Cytometry Analysis

At the end of REP, ≥ 1 × 10⁶ TIL per staining were collected and washed twice with 1× PBS. Cells were resuspended in 100 μL staining buffer (PBS + 0.5% FBS) containing flow cytometry antibodies diluted at 1:50 and stained for 30 min at 4°C in the dark. After staining, cells were washed twice with 1× PBS and analyzed using a BD FACSCanto II flow cytometer (BD Biosciences). Data were analyzed using FlowJo software (BD Biosciences). The antibodies used were FITC anti‐human CD3 (Biolegend 300,406), PE anti‐human CD4 (Biolegend 317,410), PerCP anti‐human CD8 (Biolegend 344,708), FITC anti‐human CD62L (Biolegend 304,803), APC anti‐human CD279 (PD‐1) (Biolegend 329,908), and PE anti‐human CD39 (Biolegend 328,208).

Peripheral blood was collected from patients at different time points, and red blood cells were lysed using RBC Lysis Buffer (Solarbio R1010). After lysis, samples were washed, stained, and analyzed following the same procedure as described for TIL. The antibodies used were FITC anti‐human CD3 (Biolegend 300,406), PE anti‐human CD4 (Biolegend 317,410), PerCP anti‐human CD8 (Biolegend 344,708), APC anti‐human CD25 (Biolegend 356,110), APC anti‐human CD69 (Biolegend 310,910), APC anti‐human CD279 (PD‐1) (Biolegend 329,908), and PE anti‐human CD39 (Biolegend 328,208), all diluted at 1:50 for staining.

2.6. Intracellular Cytokine/Protein Level Detection by Flow Cytometry

TIL were collected and washed twice with PBS. The cell pellet was resuspended in 100–200 μL Fixation/Permeabilization solution (BD 554714) and incubated for 30 min at 4°C in the dark. Fixed and permeabilized cells were washed twice with 1× Permeabilization/Wash Buffer. For FAM84B staining, the cell pellet was incubated with FAM84B primary antibody (Proteintech 18,421–1‐AP, diluted 1:100 in Permeabilization/Wash Buffer) for 30 min at 4°C in the dark, washed twice, and then stained with goat anti‐rabbit secondary antibody (Proteintech SA00013‐2, diluted 1:2000) for another 30 min at 4°C in the dark. Finally, cells were washed twice with Permeabilization/Wash Buffer, resuspended in FACS buffer, and analyzed using a BD FACSCanto II flow cytometer (BD Biosciences). Data were analyzed using FlowJo software.

2.7. Intracellular Cytokine Staining

Melanoma‐derived TIL were co‐cultured at 37°C with HLA‐matched melanoma cell line HMY‐1 or HLA‐mismatched A375 at an E:T ratio of 5:1. GolgiPlug (BD 555029, 1:1000) was added during stimulation. After 16 h, cells were stained with FITC anti‐human CD3 (Biolegend 300,406, diluted 1:50) to identify TIL, followed by fixation and permeabilization using the Cytofix/Cytoperm kit (BD 555028). For quantification of IFN‐γ and TNF‐α production, permeabilized cells were stained with APC anti‐human IFN‐γ (Biolegend 502,512, diluted 1:50) and PE anti‐human TNF‐α (Biolegend 502,908, diluted 1:50) for 30 min at 4°C in the dark. After staining, cells were washed twice with BD Wash Buffer and analyzed using a BD FACSCanto II flow cytometer (BD Biosciences).

2.8. Western Blot (WB) Assay

Cell lysates were prepared with RIPA buffer (20 mM NAP, pH 7.4, 150 mM NaCl, 1% Triton, 0.5% Sodium‐deoxycholate, and 0.1% SDS) containing protease inhibitor. The lysates were sonicated and centrifuged at 12,000 × g for 15 min at 4°C and subjected to protein concentration using BCA Assay Reagent (Thermo Fisher Scientific). The proteins were denatured and resolved by SDS–PAGE and immunoblotted with primary/s antibodies as follows: FAM84B (Proteintech, 18,421–1‐AP), H3 (Proteintech Cat No. 66863–1‐Ig).

2.9. HLA Typing

Peripheral blood was collected for high‐resolution HLA typing at enrollment. HLA loci were typed using a PCR sequence‐based typing method with prior DNA amplification (CapitalBio, China). Briefly, genomic DNA was extracted from peripheral blood samples according to the manufacturer's instructions of the Magic Beads DNA Extraction Kit (TANBead, China). Exons 2 and 3 of HLA class I genes (HLA‐A, HLA‐B, and HLA‐C) and exon 2 of HLA class II α and β genes (HLA‐DQ and HLA‐DR) were amplified and purified. PCR products were sequenced on an ABI 3730XL DNA Analyzer (Applied Biosystems, USA). Sequence chromatograms were analyzed using ATF1.5 software (Conexio Genomics, Australia).

2.10. Tumor Cell Lines Culture

A375 (HLA‐A*02:01) and HMY‐1 (HLA‐A*24:02) cell lines were obtained from the American Type Culture Collection (ATCC). Tumor cell lines were maintained in RPMI 1640 medium (HyClone, Cat. No. 30809) supplemented with 10% fetal bovine serum (HyClone, Cat. No. SH30406.06) and 1% penicillin–streptomycin (Gibco, Cat. No. 15140012).

2.11. In Vitro Cytotoxicity Assay

A flow cytometry‐based assay was employed to evaluate the in vitro cytotoxic activity of melanoma‐derived tumor‐infiltrating lymphocytes (TIL) against HLA‐matched melanoma cell lines. Target cells (HMY‐1) were labeled with 500 nM CFSE and seeded into 96‐well plates (50,000 tumor cells per well), followed by co‐culture with effector cells at effector‐to‐target ratios of 40:1 or 20:1 in AIM‐V medium for 16 h. Prior to analysis, cells were harvested and stained with 1 μg/mL 7‐amino‐actinomycin D (7‐AAD; BioLegend, Cat. No. 420403) for 20 min at 4°C in the dark, then analyzed by flow cytometry. The percentage of 7‐AAD‐positive cells within the CFSE‐positive population was calculated and reported as specific cytolytic lysis (%) after subtraction of the non‐effector control value.

2.12. Bulk RNA Sequencing

Bulk RNA sequencing was performed on infused TIL from all patients. mRNA was purified from total RNA using poly‐T oligo‐attached magnetic beads, followed by fragmentation into 300–350 bp fragments. First‐strand cDNA was synthesized using fragmented RNA, dNTPs (dATP, dTTP, dCTP, dGTP), and reverse transcriptase, followed by second‐strand cDNA synthesis. Double‐stranded cDNA overhangs were converted into blunt ends using exonuclease and polymerase activities. After 3′ adenylation of DNA fragments, sequencing adaptors were ligated, and library fragments were purified. The cDNA library was enriched by PCR amplification, and the PCR product was purified to generate the final library. Library concentration was quantified using the Qubit fluorometer. The precise concentration was further validated by qPCR, and fragment size distribution was assessed by agarose gel electrophoresis. Following library preparation and pooling of indexed samples, sequencing was performed on an Illumina platform. Paired‐end 150 bp (PE150) sequencing was used, as commonly applied in lncRNA‐seq.

Raw sequencing data in FASTQ format were initially processed using in‐house Perl scripts. Clean reads were generated by removing: (1) reads containing adapter sequences; (2) reads with more than 3 N bases; (3) reads with more than 20% of bases having a Qphred quality score ≤ 5. Q20, Q30, and GC content were calculated for the cleaned data. Clean reads were then mapped to the SILVA database to filter out ribosomal RNA (rRNA) sequences. All downstream analyses were conducted using rRNA‐depleted clean reads. Paired‐end clean reads were aligned to the reference genome using HISAT2. Read counts mapped to each gene were determined using FeatureCounts. Differential expression analysis was performed using EdgeR. p‐values were adjusted using the Benjamini–Hochberg method to control the false discovery rate (FDR). Genes with |log2(fold change)| > 1 and q‐value < 0.05 were considered differentially expressed.

2.13. Sanger Sequencing

Genomic DNA was extracted from the harvested cells (72 h after gRNA electroporation) using a commercial DNA extraction kit according to the manufacturer's instructions. The genomic regions flanking the CRISPR‐Cas9 target sites were amplified via polymerase chain reaction (PCR) using specifically designed primer pairs (forward and reverse primers). The PCR products were subsequently purified and verified by agarose gel electrophoresis. To evaluate the genome editing efficiency, the purified PCR amplicons (approximately 200 ng) were subjected to a reannealing process to form heteroduplex DNA. The reannealed products were then digested with T7 Endonuclease I (T7E1) at 37°C for 15–30 min. The digested fragments were separated by agarose gel electrophoresis, and the cleavage bands were visualized using a gel imaging system. To further validate the specific mutations at the target loci, the PCR products were subjected to bidirectional Sanger sequencing. The resulting sequencing chromatograms were aligned and compared with the wild‐type sequences using ICE analysis. The presence of overlapping peaks or indels (insertions/deletions) at the predicted cleavage sites (approximately 3 bp upstream of the PAM sequence) was assessed to confirm the successful introduction of targeted mutations.

2.14. Anti‐Tumor Activity of TIL in PDX Mice

Immunodeficient NTG (NOD‐scid Il2rg−/−) mice (6–8 weeks of age) were obtained from SPF Biotech Co. Ltd. (Beijing, China). A human cervical tumor was implanted into NTG mice to establish a patient‐derived xenograft (PDX) mouse model. A total of 2 × 10⁷ unedited control or gene‐edited tumor‐infiltrating lymphocytes (TIL) in saline were administered intravenously into NTG PDX mice bearing tumors with volumes of 50–100 mm³, while mice receiving saline alone served as controls. Recombinant human IL‐2 (10,000 U) was administered daily to each mouse for the first 7 days post‐TIL infusion to support the survival, proliferation, and functional activity of the transferred TIL. Body weight and tumor volume were monitored twice weekly. Approximately 0.3 mL of blood was collected from each mouse at Week 2 and Week 6 after TIL infusion. Peripheral blood mononuclear cells (PBMCs) were isolated from these samples, stained with anti‐human CD45 antibody, and analyzed by flow cytometry to detect the presence of infused TIL in peripheral circulation.

2.15. Statistical Analysis

All statistical analyses were performed using GraphPad Prism 8.0 or 9.0 (GraphPad Software, La Jolla, CA, USA). The two‐tailed Student's t‐test, Mann–Whitney U test, two‐way analysis of variance (ANOVA) with multiple comparisons adjusted by Dunnett's test, or one‐way ANOVA was used as appropriate to determine statistical significance between groups. A p‐value of ≤ 0.05 was considered statistically significant.

3. Results

3.1. Transcriptomic Analysis Identifies Potential Candidate Regulators of TIL

To preliminarily explore genes associated with TIL efficacy in cancer patients, bulk RNA sequencing was performed on infused TIL from all patients enrolled in a Phase I investigator‐initiated trial evaluating the safety and efficacy of TIL therapy in solid tumors. Basic clinical characteristics of the patients were shown in Supplemental Table 1. Exploratory transcriptomic profiles of infused TIL from Responders (R), including 3 partial responders (PR) and 1 complete responder (CR), were compared with those from Non‐responders (NR), including 5 patients with stable disease (SD) and 3 with progressive disease (PD) (Figure 2A). Differential gene expression hypothesis analysis was conducted to identify potential regulators of TIL functionality (Figure 2B). Unsupervised hierarchical clustering based on the most significantly differentially expressed genes (DEGs) revealed distinct transcriptomic signatures between the R and NR groups (Figure 2C). Among the genes significantly downregulated in the R group, ASS1, CEP20, FAM84B (LRATD2), CCR8, CTH, THBS4, TGOLN2, VSTM4, NRN1, CSF2RB, TADD, ASNS, CBS, CCR2, PXDN, FOS, and FOSB—selected based on top fold change, lowest p‐value, or known relevance to tumor immunology—were prioritized as potential candidate targets for further validation of their role in modulating TIL antitumor function. As positive controls, previously reported negative regulators of T cell antitumor activity, including Regnase‐1 (also known as ZC3H12A) [25], SOCS1 [19], and ASXL1 [26], were also included for validation.

FIGURE 2.

Molecular targets identified by transcriptomic comparison of infused TIL from responders (R) and non‐responders (NR). (A) Transcriptomic comparison between responders (n = 4) and non‐responders (n = 8) post tumor infiltration lymphocytes (TIL) therapy (Trial registration: NCT05366478, Registered on August 01, 2022, trial data not published). (B) Volcano plot of the top 46 differentially expressed genes (DEGs) of infused TIL between R and NR. A p‐value of 0.05 and log FC > 0.5 were used for the significance thresholds. (C) Heatmap of DEGs of infused TIL between R and NR, along with three control genes (ASXL1, SOCS1, ZC3H12A/Regnase‐1). (D) Current gene editing approaches and comparisons of molecular editing mechanisms of Cas9, cytosine base editors (CBE), and adenine base editors (ABE). TIL, tumor‐infiltrating lymphocytes.

3.2. Development of T‐Editor Platform

Candidate targets selected were downregulated in the R group; we hypothesized that these genes may negatively regulate the antitumor function of tumor‐infiltrating lymphocytes (TIL). To validate their functional roles, we developed a streamlined approach termed the T‐Editor platform, which leverages CRISPR/Cas9‐mediated gene editing to generate TIL with specific gene knockout (KO) or knock‐in (Figure 2D and Figure 3A). Briefly, freshly resected tumor tissues were fragmented and cultured in medium supplemented with 6,000 IU/mL IL‐2 for the pre‐rapid expansion protocol (pre‐REP). Following pre‐REP, TIL were harvested and electroporated with CRISPR/Cas9 ribonucleoprotein (RNP) complexes. After a brief recovery period, electroporated TIL were expanded in medium containing anti‐CD3 antibody, irradiated feeder cells, and 6,000 IU/mL IL‐2 for a 14‐day rapid expansion protocol (REP). At the end of REP, TIL were collected and resuspended in infusion medium to generate the final cellular product for clinical infusion (Figure 3A). The T‐Editor platform, particularly the electroporation step, was optimized by systematically evaluating multiple parameters, including pre‐electroporation stimulation conditions, electroporation programs, and CRISPR/Cas9 reagent dosing. A guide RNA (gRNA) targeting exon 3 of TGFBR2 with high specificity was used during process development.

FIGURE 3.

Development and optimization of the T‐Editor platform. (A) Schematic workflow of the T‐Editor platform showing that genetic editing was conducted between the time interval of pre‐REP and REP. (B) Detailed stimulation conditions of TIL prior to electroporation were set to generate appropriate stimulation conditions. (C) Expansion fold on REP Day 14 of NE ctrl and gene‐edited TIL under various stimulation conditions revealed that both non‐edited (NE) and gene‐edited TIL under all stimulation conditions exhibited markedly reduced expansion compared to their corresponding no stimulation group. TIL without pre‐electroporation stimulation (no stimulation group) achieved the highest expansion fold (p = 0.003 for NE and p < 0.001 for gene‐edited TIL). (D) Gene editing efficiencies of TIL under various stimulation conditions showing that editing efficiency was significantly reduced in TIL subjected to strong stimulation (stimulation‐1) prior to electroporation (p < 0.001), whereas no substantial differences were observed among the medium (stimulation‐2), weak (stimulation‐3), and no stimulation groups (p < 0.05). (E) Cell loss percentage of TIL following electroporation using MaxCyte programs: Expanding T Cell‐1 (ETC‐1), ETC‐2, ETC‐3, Resting T Cell‐14.1 (RTC‐14.1), and Optimization‐9. Among all the groups, both ETC‐1 and Optimization‐9 had the lowest cell loss (p < 0.001, p < 0.001). (F) Expansion fold on REP Day 14 of TIL electroporated using various programs. TIL of ETC‐1 and Optimization‐9 groups also displayed the highest expansion fold, which was not significantly different from the NE group (p = 0.0585, p = 0.0842), while the other 3 groups significantly reduced TIL expansion (all p < 0.01). (G) Gene editing efficiencies of TIL electroporated using various programs. Optimization‐9 group yielded the highest gene editing efficiency among all groups (p < 0.001). (H) Doses of sgRNA and Cas9 used for gene editing. (I) Expansion fold on REP Day 14 of NE ctrl and gene‐edited TIL using varying doses of sgRNA and Cas9. While the TIL expansion folds of combinations 1–4 were not significantly different from NE (all p > 0.05), combinations 5 & 6 significantly impaired TIL expansion post‐editing (p = 0.0005, p < 0.0001). (J) Gene editing efficiencies of TIL using varying doses of sgRNA and Cas9. Combinations 4–6 had similar gene editing efficiency, which was much higher than the other 3 doses (p = 0.0018, 0.0020 and 0.0021, respectively). TIL, tumor‐infiltrating lymphocytes; REP, rapid expansion protocol. The specific electroporation conditions used in Figure 2C,D are shown in Figure 2B. A two‐tailed unpaired Student's t‐test was used for comparisons between two groups. One‐way ANOVA and two‐way ANOVA followed by Tukey's multiple comparisons test were used to determine the statistical significance of three or more groups. Data are presented as mean ± SEM; a value of p < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Four pre‐electroporation stimulation conditions—strong (stimulation‐1), medium (stimulation‐2), weak (stimulation‐3), and no stimulation—were evaluated (Figure 3B). Both non‐edited (NE) and gene‐edited TIL under all stimulation conditions exhibited markedly reduced expansion compared to their corresponding no stimulation group. TIL without pre‐electroporation stimulation (no stimulation group) achieved the highest expansion fold (p = 0.003 for NE and p < 0.001 for gene‐edited TIL, Figure 3C). Gene editing efficiency was significantly reduced in TIL subjected to strong stimulation (stimulation‐1) prior to electroporation (p < 0.001), whereas no substantial differences were observed among the medium (stimulation‐2), weak (stimulation‐3), and no stimulation groups (p > 0.05, Figure 3D). These findings indicate that pre‐electroporation stimulation is not required for efficient gene editing in TIL and may impair their expansion during REP. Notably, a recent study by Dr. Jennifer Doudna's group reported that pre‐editing stimulation significantly increases the risk of chromosomal translocations in CAR‐T cell manufacturing [27]. Therefore, in the T‐Editor platform, TIL were not pre‐electroporation stimulated between pre‐REP and REP.

Electroporation was performed using the MaxCyte electroporation system, which features predefined programs, each with a unique combination of pulse duration, voltage, and waveform. MAxCyte programs Expanding T Cell‐1 (ETC‐1), ETC‐2, ETC‐3, Resting T Cell‐14.1 (RTC‐14.1) and Optimization‐9 were compared. Cell loss rate was first compared among all groups; both ETC‐1 and Optimization‐9 had the lowest cell loss (p < 0.001, p < 0.001, Figure 3E). TIL of ETC‐1 and Optimization‐9 groups also displayed the highest expansion fold, which was not significantly different from the NE group (p = 0.0585, p = 0.0842, Figure 3F), while the other 3 groups significantly reduced TIL expansion (all p < 0.01). Gene editing efficiency was also compared, and the Optimization‐9 group yielded the highest gene editing efficiency among all groups (p < 0.001, Figure 3G). Taking cell loss rate, TIL expansion, and gene editing efficiency into account, Optimization‐9 was the best electroporation program for the T‐Editor platform. Subsequently, the doses of sgRNA and Cas9 protein were optimized by testing six different concentrations (Figure 3H). While the TIL expansion folds of combinations 1–4 were not significantly different from NE (all p > 0.05), combinations 5 & 6 significantly impaired TIL expansion post‐editing (p = 0.0005, p < 0.0001, Figure 3I). Combinations 4–6 had similar gene editing efficiency, which was much higher than the other 3 doses (p = 0.0018, 0.0020 and 0.0021, respectively, Figure 3J). Considering both TIL expansion and gene editing efficiency, a dose of 0.4 nmol sgRNA and 0.2 nmol Cas9 protein (combination 4) had the highest editing efficiency without significantly impairing TIL expansion. Thus, this dose was selected for the T‐Editor platform.

3.3. Candidate Targets Screening Through T‐Editor Platform

Using the T‐Editor platform, the potential candidate target genes were individually knocked out in tumor‐infiltrating lymphocytes (TIL) for functional validation. Knockout (KO) of most genes did not affect or only mildly inhibited TIL expansion. For the majority of candidate genes, gene editing efficiency exceeded 70% (Table 1). However, SLC7A11 KO nearly abolished TIL expansion (p = 0.0005). KO of NRN1, ASNS, CBS, CCR2, PXDN, FOS, or FOSB also substantially impaired TIL expansion (p = 0.0017, p = 0.0453, p = 0.0055, p = 0.0056, p = 0.0195, p = 0.0143, p = 0.0116, Figure 4A and Table 1), suggesting these genes may be essential for T cell proliferation. Intracellular staining for IFN‐γ and TNF‐α was performed in non‐edited (NE) control and each KO TIL following polyclonal stimulation. TIL with KO of FAM84B, ASS1, CEP20, THBS4, TGOLN2, SOCS1, or Regnase‐1 exhibited more than a 1.5‐fold increase in the proportion of IFN‐γ + or TNF‐α + cells compared to NE controls (p = 0.0022, p = 0.0009, p = 0.01, p = 0.0357, p = 0.0009, p = 0.0107, p = 0.0029, respectively, Figure 4B). Cytolytic activity of control and KO TIL against HLA‐matched tumor cells was subsequently evaluated. ASNS, CBS, CCR2, PXDN, FOS, and FOSB KO TIL were excluded from the cytolysis assay due to impaired expansion and lack of enhanced cytokine production. Only FAM84B, ASS1, CEP20, and Regnase‐1 KO TIL demonstrated markedly enhanced cytolytic activity relative to NE controls (p = 0.0001, p = 0.0003, p = 0.0157, p = 0.027, Figure 4C). Collectively, these findings indicate that FAM84B, ASS1, CEP20, and Regnase‐1 might serve as critical negative regulators of TIL antitumor function. Among these four targets, FAM84B (also known as LRATD2) emerged as the top candidate, as its KO resulted in the most pronounced enhancement of TIL cytolytic activity (p = 0.0001).

TABLE 1.

Gene editing efficiency and expansion fold of DEG KO TIL.

Sample Name	Mean Gene Editing Efficiency	Mean REP D14 Expansion Fold
NE ctrl	NA	1098.9
FAM84B KO	85%	815.1
CEP20 KO	71%	606.0
ASS1 KO	75%	684.3
CCR8 KO	75%	631.1
CTH KO	90%	781.5
THBS4 KO	72%	664.1
SLC7A11 KO	60%	80.3
TGOLN2 KO	71%	809.3
VSTM4 KO	69%	634.3
NRN1 KO	83%	278.2
CSF2RB KO	77%	795.9
TRADD KO	98%	862.2
ASNS KO	78%	462.2
CBS KO	67%	433.4
CCR2 KO	75%	490.9
SOCS1 KO	93%	535.7
Regnase‐1 KO	89%	992.2
ASXL1 KO	83%	1093.3
PXDN KO	69%	499.2
FOS KO	71%	468.2
FOSB KO	78%	454.4

Abbreviations: IFN‐γ, interferon‐gamma; REP, rapid expansion; TIL, tumor‐infiltrating lymphocytes; TNF‐α, tumor necrosis factor‐α.

FIGURE 4.

Screening of candidate targets enhancing the antitumor function of TIL. (A) Expansion fold of control or gene‐edited TIL during REP. SLC7A11 KO nearly abolished TIL expansion (p = 0.0005). KO of NRN1, ASNS, CBS, CCR2, PXDN, FOS, or FOSB also substantially impaired TIL expansion (p = 0.0017, p = 0.0453, p = 0.0055, p = 0.0056, p = 0.0195, p = 0.0143, p = 0.0116), suggesting these genes may be essential for T cell proliferation. (B) Percentage of IFN‐γ‐ or TNF‐α‐producing cells following stimulation, measured by intracellular cytokine staining. TIL with KO of FAM84B, ASS1, CEP20, THBS4, TGOLN2, SOCS1, or Regnase‐1 exhibited profound increase in the proportion of IFN‐γ + or TNF‐α + cells compared to NE controls (p = 0.0022, p = 0.0009, p = 0.01, p = 0.0357, p = 0.0009, p = 0.0107, p = 0.0029). Final conditions selected for this assay: TIL are not pre‐stimulated before electroporation; 0.4 nmol sgRNA and 0.2 nmol Cas9 protein are the final doses used. (C) In vitro cytolysis assay of melanoma‐derived, HLA‐A*24:02‐restricted control and edited TIL using HLA‐matched melanoma cells as targets at various effector‐to‐target (E:T) ratios. FAM84B, ASS1, CEP20, and Regnase‐1 KO TIL demonstrated markedly enhanced cytolytic activity relative to NE controls (p = 0.0001, p = 0.0003, p = 0.0157, p = 0.027). TIL, tumor‐infiltrating lymphocytes; IFN‐γ, interferon‐gamma; TNF‐α, tumor necrosis factor‐α; REP, rapid expansion protocol. Two‐tailed unpaired Student's t‐test was used for comparisons between two groups. One‐way ANOVA and two‐way ANOVA followed by Tukey's multiple comparisons test were used to determine the statistical significance of three or more groups. Data are presented as mean ± SEM; a value of p < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

3.4. Base Editing of FAM84B Demonstrates High Manufacturing Efficiency and Feasibility

Given that FAM84B was identified as the most influential target on TIL anti‐tumor activity among the genes we tested, we aimed to develop a novel function‐enhanced TIL therapy using the T‐Editor platform to disrupt FAM84B. However, Cas9‐mediated double‐strand breaks (DSBs) in DNA are a known cause of chromosomal translocations [28, 29], potentially increasing the tumorigenic risk of edited TIL and limiting clinical translation. In contrast, base editing enables direct, programmable conversion of one base pair to another without inducing DSBs [30]. AccuBase is a cytosine base editor (CBE) that efficiently and precisely converts C•G to T•A with low off‐target activity [31, 32]. A CAR‐NK cell therapy utilizing AccuBase has received FDA IND approval [32]. Since CBE is a Cas9‐derived fusion protein that, like Cas9, forms ribonucleoprotein (RNP) complexes with guide RNA (gRNA), we hypothesized that FAM84B base editing in TIL could be efficiently and feasibly achieved using the T‐Editor platform. Two gRNAs were designed for Cas9‐mediated KO and two for AccuBase‐mediated base editing (Figure 5A). All gRNAs targeted exon 2 of the FAM84B gene to ensure premature termination of translation and effective functional knockout. FAM84B gRNA1 and gRNA2 induced indel mutations with editing efficiencies of 85% and 79%, respectively (Figure 5B,C). With FAM84B gRNA3, the fifth C•G base pair within the spacer sequence was converted to T•A at a rate of 73%, generating a TAG stop codon and resulting in early translational termination and gene knockout. Base editing was predominantly confined to the intended site (fifth C•G), with minimal off‐target editing at the fourth C•G (12% conversion) (Figure 4D). In contrast, FAM84B gRNA4 achieved only 32% editing at the intended third C•G site and exhibited frequent editing at the second, fourth, fifth, and sixth C•G positions (Figure 5E). Thus, gRNA3 mediated more precise and focused base editing than gRNA4, with knockout efficiency comparable to that of gRNA1 and gRNA2. At the protein level, FAM84B expression was markedly reduced following gRNA3‐mediated base editing (p = 0.005, Figure 5F). Western blot assay further confirmed the significant reduction of FAM84B protein expression following editing (p = 0.004 and p = 0.026, respectively, Figure 5G,H).

FIGURE 5.

Selection of gRNAs for FAM84B editing. (A) Genomic locations of sgRNAs targeting FAM84B. (B, C) Gene editing efficiencies of gRNA1/gRNA2 for Cas9‐mediated FAM84B knockout were 85% and 79%, respectively. The electrophoresis quantification map of T7E1 enzyme cleavage product on the top, showing the cleavage efficiency and the ratio of knockouts (the blue column). The original chromatogram and sequence alignment of Sanger sequencing of the target region are shown in the middle, displaying obvious overlapping peaks and Indel mutations. At the bottom of the figures, the mutated area has been enlarged, indicating that the mutation actually exists. (D) Base editing efficiency of gRNA3. The fifth C•G base pair within the spacer sequence was converted to T•A at a rate of 73%, generating a TAG stop codon and resulting in early translational termination and gene knockout. Base editing was predominantly confined to the intended site (fifth C•G), with minimal off‐target editing at the fourth C•G (12% conversion). (E) Base editing efficiency of gRNA4. This gRNA achieved only 32% editing at the intended third C•G site and exhibited frequent editing at the second, fourth, fifth, and sixth C•G positions. (F) Reduction in FAM84B protein levels following base editing with gRNA3. FAM84B expression was markedly reduced following gRNA3‐mediated base editing compared with the control of secondary‐only staining (p = 0.005). (G) Western blot images of FAM84B protein levels in control and FAM84B KO TIL. (H) Quantification of relative densities in the control and FAM84B KO TIL, which showed that there were significant reductions of FAM84B protein expression following editing (p = 0.004 and p = 0.026, respectively). Flow cytometry and Western blot analysis were performed 72 h post‐genetic editing. Two‐tailed unpaired Student's t‐test was used for comparisons between two groups. One‐way ANOVA and two‐way ANOVA followed by Tukey's multiple comparisons test were used to determine the statistical significance of three or more groups. Data are presented as mean ± SEM; a value of p < 0.05 was considered significant.

Both FAM84B knockout TIL generated with gRNA1 and CBE‐edited TIL generated with gRNA3 expanded at a rate comparable to each other (p = 0.9406), albeit slightly lower than that of non‐edited control TIL (p = 0.0767, Figure 6A), suggesting that FAM84B is not essential for TIL proliferation. The exhaustion and memory phenotypes of the edited TIL were also similar to controls, as evidenced by comparable proportions of CD39 + PD‐1+, CD62L+, and CD45RO+ populations. The frequency of CD39 + PD‐1+ cells in both FAM84B loss‐of‐function TIL populations was similar to that in non‐edited controls, indicating that FAM84B does not regulate TIL exhaustion during ex vivo expansion (Figure 6B). However, CD62L+ naïve‐like memory T cells were more enriched in FAM84B‐deficient TIL than in control TIL, suggesting a repressive role of FAM84B in maintaining the memory phenotype (p = 0.0005, p = 0.0226, Figure 6B,C). Flow cytometry analysis of central memory T cells (CD62L + CCR7+) and terminal effector memory T cells (CD62L‐CCR7‐CD127+) revealed a significant increasing of CD62L + CCR7+ T and CD62L‐CCR7‐CD127 + T cells post FAM84B knocking out (p = 0.0034 and p = 0.0037, respectively, Figure 6D,E). There were no significant differences in classic T cell exhaustion (TIM3 + PD‐1+) and severe T cell exhaustion (TIM3 + PD‐1 + CD39+) between the control and FAM84B knockout TIL (p > 0.05, Figure 6F,G). The results further confirmed the significantly increased memory‐like population of FAM84B KO TIL.

FIGURE 6.

Expansion and phenotypic characterization of FAM84B base‐edited TIL. (A) Expansion fold of control TIL, FAM84B knockout TIL (mediated by Cas9/gRNA2), and FAM84B base‐edited TIL (mediated by CBE/gRNA3) during REP. Both FAM84B knockout TIL generated with gRNA1 and CBE‐edited TIL generated with gRNA3 expanded at a rate comparable to each other (p = 0.9406), albeit slightly lower than that of control TIL (p = 0.0767). (B) Exhaustion and memory phenotypes of control, FAM84B KO, and FAM84B CBE TIL. (C) Quantitative representation of exhaustion and memory phenotypes in bar graphs. The frequency of CD39⁺PD‐1⁺ cells in both FAM84B loss‐of‐function TIL populations was similar to that in non‐edited controls. CD62L⁺ memory T cells were more enriched in FAM84B‐deficient TIL than in control TIL (p = 0.0005, p = 0.0226). TIL, tumor‐infiltrating lymphocytes. (D, E) Flow cytometry analysis of central memory T cells (CD62L + CCR7+) and terminal effector memory T cells (CD62L‐CCR7‐CD127+), and there was a significant increase in CD62L + CCR7+ and CD62L‐CCR7‐CD127+ cells post‐FAM84B knockout (p < 0.05). (F, G) Flow cytometry analysis of TIM3 + PD‐1+ to indicate classic T cell exhaustion and TIM3 + PD‐1 + CD39+ for severe T cell exhaustion. There was no significant difference in T cell exhaustion between the control and FAM84B knockout TIL. Two‐tailed unpaired Student's t‐test was used for comparisons between two groups. One‐way ANOVA and two‐way ANOVA followed by Tukey's multiple comparisons test were used to determine the statistical significance of three or more groups. Data are presented as mean ± SEM; a value of p < 0.05 was considered significant.

3.5. FAM84B‐Edited TIL Generated via the T‐Editor Platform Represents a Potential Novel and Enhanced TIL Product

To evaluate the in vitro anti‐tumor activity of FAM84B knockout TIL (gRNA1) and CBE‐edited TIL (gRNA3), we employed the allogeneic melanoma cell line HMY‐2, which expresses the HLA‐A*24:02 haplotype. Both edited and non‐edited control TIL were derived from the same HLA‐A*24:02‐positive melanoma tissue and co‐cultured with HMY‐2 cells. Following co‐culture, the concentrations of IFN‐γ and TNF‐α in the supernatants of Cas9‐edited and CBE‐edited TIL were 5‐fold and 4‐fold higher, respectively, compared to control TIL (p = 0.0009, p = 0.0002, p = 0.0006, p = 0.0038, Figure 7A). Both edited TIL populations exhibited significantly greater in vitro cytolytic activity against HMY‐2 cells than control TIL (p = 0.0157, p = 0.0044, Figure 7B). The cytokine production and cytotoxic potency of Cas9‐edited TIL were comparable to those of CBE‐edited TIL. To substantiate that the observed reactivity is HLA‐A24:02‐restricted, not an enhancement of a non‐specific T cell reactivity, we performed an MHC Class I blocking assay using the W6/32 antibody. The result revealed that there was a significant decrease in cytotoxic lysis when the MHC Class I was blocked with the W6/32 antibody in both control and FAM84B KO TIL groups (p < 0.001, p < 0.001, Figure 7C). There was no significant difference between the control and FAM84B KO TIL when blocking with the MHC Class I molecule (p = 0.1907, Figure 7C), implying that the observed results were MHC Class I restricted in this experiment.

FIGURE 7.

In vitro and in vivo evaluation of FAM84B base‐edited TIL demonstrating enhanced antitumor responses. (A) Concentrations of IFN‐γ and TNF‐α in culture supernatants of control and edited TIL following antigen stimulation. The concentrations of IFN‐γ and TNF‐α of Cas9‐edited and CBE‐edited TIL were 5‐fold and 4‐fold higher, respectively, compared to control TIL (p = 0.0009, p = 0.0002, p = 0.0006, p = 0.0038). (B) In vitro cytolysis assay of melanoma‐derived, HLA‐A*24:02‐restricted control and edited TIL using HLA‐matched melanoma cells as targets. Both edited TIL populations exhibited significantly greater in vitro cytolytic activity against HMY‐2 cells than control TIL (p = 0.0157, p = 0.0044). (C) MHC Class I blocking assay showed that there was a significant decrease in cytotoxic lysis of both the control and FAM84B KO TIL when we blocked the MHC Class I with the W6/32 antibody with an effector/TIL ratio (E/T) of 20:1(p < 0.001). W6/32 = 20ug/ml. There was no significant difference between the control and FAM84B KO TIL by blocking with the MHC Class I molecule (p = 0.1907). (D) Study design for in vivo efficacy assessment using an HLA‐A*24:02‐restricted patient‐derived xenograft (PDX) mouse model. Due to unavailability of autologous TIL for this PDX model, infused TIL were derived from an HLA‐A*24:02‐matched donor. (E) Tumor volume measurements in PDX mice (n = 6). Control TIL achieved a tumor growth inhibition (TGI) of 7% on day 23 post‐infusion, whereas FAM84B CBE‐edited TIL demonstrated markedly enhanced anti‐tumor efficacy, with a TGI of 28% (p = 0.0009). (F) Percentage of human CD45+ CD3+ cells in peripheral blood from mice in each group. Mice infused with FAM84B CBE‐edited TIL exhibited higher percentages of human CD45⁺CD3⁺ cells in peripheral blood at multiple time points compared to those receiving control TIL (p < 0.0001). (G) Serum IFN‐γ levels in peripheral blood from mice in each group. Mice infused with FAM84B CBE‐edited TIL exhibited elevated IFN‐γ levels in peripheral blood at multiple time points compared to those receiving control TIL (p < 0.0001). TIL, tumor‐infiltrating lymphocytes; IFN‐γ, interferon‐gamma; TNF‐α, tumor necrosis factor‐α. Two‐tailed unpaired Student's t‐test was used for comparisons between two groups. One‐way ANOVA and two‐way ANOVA followed by Tukey's multiple comparisons test were used to determine the statistical significance of three or more groups. Data are presented as mean ± SEM; a value of p < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

For the in vivo efficacy assessment, a patient‐derived cervical tumor expressing HLA‐A*24:02 was engrafted into severely immunodeficient mice to establish a PDX model. Due to the unavailability of autologous TIL for this model, both control and edited TIL were derived from an allogeneic HLA‐A*24:02‐positive cervical tumor. Vehicle, control TIL, or FAM84B CBE‐edited TIL (gRNA3) were infused into tumor‐bearing mice, followed by administration of supportive IL‐2 for 7 days (Figure 7D). Control TIL exerted moderate anti‐tumor activity, achieving a tumor growth inhibition (TGI) of 7% on day 23 post‐infusion. In contrast, FAM84B CBE‐edited TIL demonstrated markedly enhanced anti‐tumor efficacy, with a TGI of 28% (p = 0.0009, Figure 7E), indicating that FAM84B negatively regulates the in vivo antitumor function of TIL. Furthermore, mice receiving FAM84B CBE‐edited TIL exhibited higher percentages of human CD45 + CD3+ cells and elevated IFN‐γ levels in peripheral blood at multiple time points compared to those receiving control TIL, indicating improved TIL persistence and a stronger systemic immune response (p < 0.0001, p < 0.0001, Figure 7F,G). In summary, FAM84B possibly acts as a negative regulator of TIL‐mediated anti‐tumor efficacy both in vitro and in vivo. CBE‐mediated loss‐of‐function of FAM84B in TIL generates a novel and more potent TIL product with significantly enhanced therapeutic potential.

4. Discussion

Although tumor‐infiltrating lymphocyte (TIL) therapy has emerged as a promising approach for patients with advanced solid tumors refractory to standard‐of‐care treatments, clinical responses remain modest. Objective response rates to conventional non‐engineered TIL therapy range from 25% to 45%, depending on tumor type, and a significant proportion of patients experience disease progression and develop resistance within months after TIL infusion. The underlying mechanisms and factors influencing therapeutic response are not well understood. Given that TIL products are autologous, individualized mixtures of T cells expressing thousands of T cell receptors (TCRs) and existing in diverse functional states, increasing attention is being directed toward understanding the relationship between TIL product characteristics and clinical outcomes [10, 11]. Identifying intrinsic TIL factors that determine clinical efficacy may facilitate optimization of TIL manufacturing processes and inform the development of next‐generation gene‐engineered TIL therapies, ultimately improving response rates.

To address this knowledge gap, our study focused on exploring the association between the transcriptomic profile of TIL products and clinical response data from a Phase I investigator‐initiated trial evaluating the safety and efficacy of TIL therapy in solid tumors. Seventeen differentially expressed genes (DEGs) were identified in TIL from responders (R) vs. non‐responders (NR) as possible candidate mediators of therapeutic efficacy. However, the functional validation of these potential candidate genes was hindered by the lack of simple and efficient methods to genetically engineer TIL, which are often exhausted, highly differentiated, and exhibit relatively low viability. To overcome this limitation, we developed a streamlined, robust, and reliable genetic engineering platform for TIL—termed T‐Editor. This platform enables rapid and efficient CRISPR‐mediated gene editing by incorporating only minimal additional steps (1.5–2 h) into the standard TIL manufacturing process. Key parameters—including stimulation conditions, electroporation protocols, and CRISPR/Cas9 reagent dosing—were systematically optimized for the T‐Editor platform.

The T‐Editor platform has two distinguishing features. First, it eliminates the need for pre‐engineering T cell stimulation. Such stimulation is commonly employed in the manufacturing of CAR‐T and TCR‐T cells (e.g., for viral transduction or gene editing) [33, 34, 35], likely due to the high proportion of naïve T cells in PBMCs, which reside in a quiescent state characterized by condensed chromatin and low transcriptional and translational activity—factors that impede genetic modification. Although most published studies on TIL genetic manipulation also include a pre‐stimulation step [20, 21, 22, 23], the majority of tumor‐infiltrating T cells are already highly differentiated. We compared multiple stimulation conditions and found that stimulating pre‐REP TIL prior to electroporation did not significantly enhance gene editing efficiency but instead impaired subsequent TIL expansion during REP culture. Furthermore, recent work by Dr. Jennifer Doudna's group demonstrated that pre‐editing stimulation markedly increases the risk of chromosomal translocations in CAR‐T cell manufacturing [27]. Therefore, the T‐Editor platform offers a simpler, faster, and potentially safer alternative to current mainstream TIL engineering approaches. Second, T‐Editor is a scalable technology. Electroporation is performed using the MaxCyte ATx platform, which supports a broad range of cell numbers (7.5 × 10⁴ to 2 × 10¹⁰) and is compatible with good manufacturing practice (GMP)‐grade reagents. This allows the electroporation conditions established in this study to be directly scaled up for clinical‐grade TIL gene editing.

With the aid of the T‐Editor platform, we generated tumor‐infiltrating lymphocytes (TIL) with knockout (KO) of each target gene. We established multidimensional assays—including cell expansion, cytokine production, and cytolytic activity—to compare NE control TIL with these KO TIL, and FAM84B emerged as the top candidate, as its KO most markedly enhanced TIL antitumor function. FAM84B functions as a potent oncogene primarily by promoting proliferation through key signaling pathways such as β‐catenin and NF‐κB [36, 37]. In breast cancer, it drives tumorigenesis by activating the NF‐κB and death receptor signaling pathways [37]. Although the death receptor signaling pathway is essential for Fas‐ and TNF‐mediated apoptosis—processes linked to T‐cell function—the direct role of FAM84B in regulating the anti‐tumor activity of TIL has not previously been reported. To our knowledge, this study is the first to identify a novel function of FAM84B as a possible negative regulator of TIL potency. FAM84B KO did not significantly affect TIL expansion; however, a higher proportion of CD62L+ memory T cells was observed in FAM84B‐deficient TIL compared to control TIL. Loss of FAM84B in TIL not only significantly enhanced cytokine production and cytolytic activity in vitro, but also improved their in vivo antitumor efficacy in an HLA‐matched PDX mouse model. In vivo anti‐tumor experiments play an indispensable and multi‐faceted role in the development of anti‐tumor therapies. Mouse models with human tumor xenografts are important for pre‐clinical testing of TIL therapies. This model can closely mimic the complex tumor microenvironment in humans and observe the TIL's efficacy in inhibiting tumor growth in vivo, which is of great significance in advancing our understanding of developing effective anti‐tumor therapies of TIL and ensuring the safety and efficacy of TIL in clinical practice. However, due to the unavailability of autologous TIL for this PDX model, infused TIL for the in vivo experiment were derived from an HLA‐matched donor. While it was worth noting that allogeneic TIL can potentially recognize a wide range of tumor antigens and induce an anti‐tumor immune response, and has been used in in vivo studies due to its immunological relevance and practicality. The use of HLA‐matched allogeneic TIL in the present study allows for a more realistic simulation of the immune response, which can enhance the reliability of the in vivo model in predicting the efficacy of edited TIL. Thus, both findings of the present study establish FAM84B as an attractive target for engineering next‐generation TIL therapies.

Previous studies have suggested that FAM84B is associated with the β‐catenin/Wnt signaling pathway [38, 39]. The activation of the Wnt pathway leads to the accumulation and nuclear translocation of β‐catenin, which possibly binds to transcription factor TCF1 and activates the transcription of target genes in the nucleus of TIL [38, 39]. FAM84B possibly inhibits the Wnt/β‐catenin pathway, reducing the availability of β‐catenin for TCF1 [38, 39]. This could potentially disrupt the normal gene regulatory network for memory T cell formation in edited TIL. Thus, the increased memory‐like T cells we observed in edited TIL might be due to FAM84B, through its influence on TCF1 activity via the Wnt/β‐catenin pathway, directly affecting the development and maintenance of memory T cell survival and renewal. Further in‐depth studies are required to uncover the molecular mechanisms of whether and how FAM84B regulates the survival or proliferation of memory T cells through TCF1.

This study also represents the first attempt to engineer TIL using CBEs rather than conventional Cas9, which carries a higher risk of chromosomal translocations. AccuBase, the CBE employed in this study, features an extremely narrow editing window, enabling precise discrimination between target cytosines and neighboring bystander bases, thereby greatly enhancing editing specificity. It operates via a “conditional activation” mechanism that keeps the deaminase enzyme inactive until it reaches the correct DNA target, substantially reducing off‐target genomic edits [31]. Base‐edited TIL achieved FAM84B KO efficiency comparable to that of Cas9‐edited TIL. Furthermore, the expansion capacity, phenotypic profile, cytokine production, and in vitro cytolytic activity of base‐edited TIL were similar to those of Cas9‐KO TIL, indicating that CBEs are as effective as Cas9 for TIL engineering, but with a superior safety profile due to the absence of DSBs.

Despite these promising results, this study has several limitations. First, the sample size in the Phase I trial was small, resulting in insufficient transcriptomic data from responders (R) and non‐responders (NR) for comprehensive comparison. Consequently, several key genes involved in regulating TIL antitumor function may not have been captured among the significant DEGs. Second, our target screening was limited to in vitro assays—such as cell expansion, cytokine production, and cytolytic activity—which may overlook candidate genes that influence TIL infiltration or enhance resistance to the immunosuppressive tumor microenvironment. In addition, these data were derived from limited TIL samples across various gene knockouts, and further investigations using more TIL samples are needed to improve the data generalizability. Moreover, although we demonstrated that FAM84B plays a potentially critical role in negatively regulating TIL anti‐tumor activity, the underlying molecular mechanisms remain unclear. Future mechanistic studies will be essential to support the development of next‐generation TIL therapies targeting FAM84B. It is worth noting that TIL products are highly heterogeneous; there are several aspects that can contribute to the observed transcriptomic variability. Firstly, different TCR clonotypes may recognize distinct tumor antigens, leading to unique activation and signaling pathways. As a result, each clone may link to a different set of genes, causing variations in the overall transcriptomic profile. Secondly, CD4⁺ T cells and CD8⁺ T cells are involved in helper functions, such as cytokine secretion and costimulation, and are responsible for direct cytotoxicity, respectively. CD4/CD8 composition in TIL products may result in different gene expression patterns that can lead to transcriptomic differences. Thirdly, tumor‐reactive T cells are actively engaged in the anti‐tumor response and may express genes related to activation, cytotoxicity, and immune regulation. In contrast, bystander T cells, which are not directly involved in the anti‐tumor response, may have a different gene expression profile. Lastly, naive, effector, and memory T cells have distinct gene expression patterns. As T cells differentiate, their gene expression changes, which can lead to variations in the overall transcriptomic data. Bulk RNA‐seq averages the gene expression across all cells in the sample, masking the differences between individual cell populations; thus, diverse TCR clonotypes and functional T cell states of each TIL sample may cause the overall transcriptomic variability. It is important for future studies to use single‐cell RNA‐seq to better understand the transcriptomic profiles of different TIL subpopulations.

In conclusion, we developed a rapid and efficient CRISPR‐mediated gene editing platform for TIL engineering, identified FAM84B as a novel potential target to enhance anti‐tumor activity, and pioneered the use of CBE to generate FAM84B loss‐of‐function TIL with enhanced potency. Future studies should focus on elucidating the mechanistic role of FAM84B in regulating TIL function and conducting comprehensive preclinical evaluations of the safety, biodistribution, and efficacy of FAM84B base‐edited TIL prior to initiating first‐in‐human clinical trials.

Author Contributions

Fenge Li: conceptualization, investigation, funding acquisition, writing – original draft, validation, project administration, software, supervision, resources. Xuejiao Ma: writing – original draft, investigation, validation, formal analysis, software, resources, supervision. Yongming Xue: investigation, writing – original draft, validation, visualization, formal analysis, data curation, resources, project administration. Jilong Yang: writing – review and editing, software, project administration, supervision, data curation, resources, validation, investigation. Ruijing Wang: software, resources, supervision, writing – review and editing, validation, formal analysis, investigation. Xiang Li: visualization, writing – review and editing, data curation, resources, supervision, validation, investigation. Chao Zhang: writing – review and editing, validation, formal analysis, supervision, data curation, project administration, investigation. Wenhan Lu: writing – review and editing, formal analysis, data curation, methodology, validation, investigation. Mei Liu: visualization, writing – review and editing, resources, formal analysis, project administration, validation, investigation. Yue Xu: data curation, writing – review and editing, visualization, formal analysis, software, resources, investigation. Xinyi Wang: writing – review and editing, validation, resources, software, investigation. Shengnan Wu: writing – review and editing, resources, data curation, validation. Ning Kang: writing – review and editing, validation, methodology, resources. Chunhua Ma: writing – review and editing, resources, supervision, project administration, methodology, visualization, funding acquisition, conceptualization.

Funding

This work was supported by the horizontal research fundings (Funding Nos. 2019SHHLMCH, 2024YJ023, 2019ZGJCJJLM, 2024YJ020, H, HKRA20250802, 320.6750.2025‐20‐31); the Key research project funded by the First Affiliated Hospital of Nankai University key funding project (funding no. 2025YJZD002); the Natural Science Foundation of Tianjin established by Nankai University independently (No. 24JCYBJC01590); the Tianjin Key Medical Discipline Construction Project (Grant No. TJYXZDXK‐3‐003A‐5), and the Funded by Tianjin Key Specialty Construction Project for combining traditional Chinese and Western medicine (No. ZDZKKF02).

Ethics Statement

Ethics approval: This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and was approved by the Tianjin Anti‐Cancer Association and the Ethics Committee of the First Affiliated Hospital of Nankai University. The data in Figure 1 of the Trial registration: NCT05366478, registered on August, 01 2022 (clinicaltrials.gov). Written informed consent was obtained from both patients prior to participation in the study. Animal studies were approved by the Ethics Committee of the First Affiliated Hospital of Nankai University.

Conflicts of Interest

Y.X. and W.L. are employees, and F.L. is a shareholder of Suzhou Lanma Biotechnology Co. All other authors declare no conflicts of interest. None of the authors of this manuscript is a current Editor or Editorial Board Member of Cancer Science.

Supporting information

Supplemental Table 1 Patient clinical and demographic characteristics at baseline.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.