Expansion of KRAS hotspot mutations reactive T cells from human pancreatic tumors using autologous T cells as the antigen-presenting cells

Adoptive cell therapy (ACT) with expanded tumor-infiltrating lymphocytes (TIL) or TCR gene-modified T cells (TCR-T) that recognize mutant KRAS neo-antigens can mediate tumor regression in patients with advanced pancreatic ductal adenocarcinoma (PDAC) (Tran et al in N Engl J Med, 375:2255–2262, 2016; Leidner et al in N Engl J Med, 386:2112–2119, 2022). The mutant KRAS-targeted ACT holds great potential to achieve durable clinical responses for PDAC, which has had no meaningful improvement over 40 years. However, the wide application of mutant KRAS-centric ACT is currently limited by the rarity of TIL that recognize the mutant KRAS. In addition, PDAC is generally recognized as a poorly immunogenic tumor, and TILs in PDAC are less abundant than in immunogenic tumors such as melanoma. To increase the success rate of TIL production, we adopted a well-utilized K562-based artificial APC (aAPC) that expresses 4-1BBL as the costimulatory molecules to enhance the TIL production from PDCA. However, stimulation with K562-based aAPC led to a rapid loss of specificity to mutant KRAS. To selectively expand neo-antigen-specific T cells, particularly mKRAS, from the TILs, we used tandem mini gene-modified autologous T cells (TMG-T) as the novel aAPC. Using this modified IVS protocol, we successfully generated TIL cultures specifically reactive to mKRAS (G12V). We believe that autologous TMG-T cells provide a reliable source of autologous APC to expand a rare population of neoantigen-specific T cells in TILs.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) has a 9% fiveyear survival rate, typically presented as an unresectable, locally advanced, and metastatic disease [3]. The firstline treatment for pancreatic cancer is surgery, radiation, and chemotherapy [3,4]. Nearly all patients will develop chemo-resistant tumors and progress to progressive, metastatic PDAC within two years of diagnosis. Despite the promising successes of small molecule targeted therapy and immunotherapy in several other cancer types, advance in the treatment of advanced PDAC is dismally slow [5,6]; there is no effective treatment beyond the standard but minimally effective chemotherapy regimens.
Generation of TIL culture and rapid expansion of TILs for clinical application has been pioneered and successfully developed by Rosenberg's team at Surgery Branch at the NCI to treat patients with melanoma and renal cell carcinoma (RCC) [7]. Goff et al. demonstrated that tumorreactive TIL cultures were successfully generated in most patients (94%), and tumor-specific TILs were identified in Sizhen Wang, Xiaohui Zhang, Xuemei Zou, Maorong Wen have contributed equally to this work.

3
two-thirds of patients and expanded enough to treat 27% of melanoma patients (107 out of 402 patients) [8]. PDAC is known to have an immune-suppressive microenvironment that blocks the infiltration of CD8 + T cells. The CD3 + TILs, predominantly CD4 + T reg cells, detected in either primary or metastatic pancreatic tumors were less than 1% of all cells compared to > 2% in the metastatic melanoma [9]. Sakellariou-Thompson et al. used an agonistic 4-1BB mAb to increase the success rate and yield of TILs from PDAC and preferential expansion of CD8 + T cells. Other groups take advantage of artificial APC (aAPC) engineered to express 4-1BBL or other co-stimulatory molecules in combination with cytokines to increase the proliferation of TILs [10,11]. A high concentration of recombinant human IL-2 was typically used to expand TILs; however, combined with the other common cytokine receptor γc cytokine such as IL-15 and IL-21 was shown to be beneficial for the selective expansion of antigen-specific T cells and maintaining expanded T cells in a less differentiated stage [12][13][14]. When 4-1BBLexpressing K562 cells were further engineered to produce soluble IL-21, a superior expansion of less differentiated antigen-specific T cells was observed without concomitant proliferation of regulatory T cells [15]. When further engineered to express membrane-bound IL-21, K562-based artificial 4-1BBL-expressing aAPCs were also very effective for the ex vivo expansion of NK cells from peripheral blood or cord blood [16,17]. Based on the published literature, we set out to develop a reliable in vitro TIL generation protocol from PDAC. Because more than 90% of PDAC cells harbor the KRAS mutation, we also investigated whether T cells that recognize mutant KRAS (mKRAS) in TILs could be specifically expanded via in vitro sensitization (IVS) method with aAPC or autologous PB T cells engineered to express mKRAS tandem minigenes (TMG-T).

Antibodies
The following antibodies listed were obtained from Biolegend: FITC anti-human CD3 antibody (cat. 300

Clinical samples
The PDAC tissues and blood samples were provided by the affiliated hospital of Nanjing University. Tumor tissue collection was approved by the Committee on Human Research and the institutional review board of the hospital under protocol number BTPC-01. In addition, all patients provided written, informed consent before biopsy and blood collection. Patient sample details, including the biopsy location and neo-adjuvant treatment information, are found in Table S1.

Feeder cells enhance the in vitro expansion of TILs from tumor fragment cultures.
Tumors or para-tumor tissues were dissected into small 1-3mm fragments and placed in one well of a 24-well plate and 1 mL of TIL medium. Typically, 2 × 10 5 feeder cells were added to each well. Half of the media were changed with new TIL media every other day, and the cell number was enumerated at 12-14 days.
For TMG-T generation, PBMC from HLA-A11:01 donors was pre-activated with 50 ng/ml soluble anti-CD3 antibody (OKT3) supplemented with 300 IU/ml IL-2 for two days before the retroviral transduction. Retroviral supernatant was first centrifuged for transduction onto Retronectin-coated (15 µg/ml, Takara) 24-well plates at 2000 × g for two hours at 32 °C. Pre-activated T cells (1 ml per well, at 5 × 10 5 cells/ ml in IL-2 containing T-cell media) were then spun onto the retrovirus plates for 10 min at 1000 × g. Transduced T cells were further cultured in T-cell media. TMG-T cells were sorted based on GFP expression and expanded according to the REP for 11-16 days before being frozen in liquid nitrogen. Transduced T cells were further expanded in GREX flasks (Wilson Wolf) [18,19].

Generation of T cells that recognize mutant KRAS.
The gene encoding the PCV TCR that recognizes 9-mer peptide (VVGAVGAGK) in the context of HLA-A11:01 was synthesized and cloned into the MSGV1 vector [20]. The PCV TCR was engineered to express mouse TCR's constant alpha or beta for correct pairing and easy detection with an anti-mouse TCR beta chain antibody. The retrovirus supernatant and transduction of PB T cells are as described for TMG-T cells.

Optimization of in vitro sensitization (IVS) protocol with DoE with autologous T cells that express TMG (TMG-T) cells.
To mimic TILs that contain a low frequency of mKRASspecific T cells, we diluted T cells expressing PCV TCR with untransduced T cells to yield an mTCR positive rate of 0.5%. The final concentration of human AB serum has three gradients (2%, 6%, and 10%). Mitomycin C-treated or irradiated TMG-T cells were mixed with responder T cells (2 × 10 5 ), resulting in a 1:1, 1:5.5, or 1:10 responder T-cells ratio to TMG-T cells (2 × 10 5 T cells combined with either 2 × 10 5 , 1.1 × 10 6 , or 2 × 10 6 TMG-T cells). Irradiated allogeneic PBMC (30 Gy) was added to the culture at the same proportion as TMG-T cells. The total medium volume per well was adjusted to 2 ml, containing 50 ng/ml of IL-21 and 300 IU/ml of IL-2. The IL-21 was superior to IL-7 and IL-15 combined with a complete TIL medium supplemented with 300 IU/ml of IL-2 (data not shown). Half of the media was changed every 3-4 days, and mTCR frequency was assessed using FACS after 12 days. The IVS optimization experiment via the Design of Experiment (DOE) was repeated once. We used the JUMP software to analyze the results, define the optimal parameters, and follow experimental validation.

Flow cytometry analysis of surface markers and functional degranulation assay.
On day 14 of tumor fragment culture, lymphocytes in each well were harvested and stained with antibodies conjugated with different fluorochromes, as listed in the material section. Briefly, cells were harvested from each well, washed with PBS, and blocked Fc before staining with antibody mixtures with a predetermined dilution at 4 °C. The expression of different surface markers was measured by flow cytometry analysis using the Spectral Cell Analyzer SA3800 (SONY) and analyzed using Flowjo software (Tree Star).

Generation of MHC tetramer and identification of T cells that bind HLA*A11:01 complexed with G12V peptides
Recombinant monomeric pMHCs were assembled in-house, and tetramers were prepared with modification [21]. The extracellular domain of HLA-A11 and β2-microglobulin was produced as inclusion bodies in bacteria and refolded in a solution that contained synthetic peptides according to the standard procedures, followed by purification via anion exchange and size-exclusion chromatography. After chemical biotin modification, the pMHC-containing fractions were further purified by gel filtration on a Superdex 200 column. For pMHC tetramer assembly, 100 ml of pMHC monomer (15 mM) and 100 ml SA-APC (0.2 mg/ml, Biolengend) were mixed, followed by incubation on ice for 20 min. Tetramers were stored at four °C in the dark with added 0.02% NaN 3 and typically used within three days after their assembly. The specificity of house-made tetramer was validated (supplementary data Figure S4). Approximately one million TIL cells or cells transduced with TCRs were used for tetramer staining. The cells were washed with FACS buffer (D-PBS + 2% FBS) and pretreated with the PKI (50 nM Dasatinib) for 30 min at 37 °C. We washed cells and incubated them with the Zombie Green viability dye (BioLegend, 65-0866-14) for 10 min at room temperature in the dark. After incubation, 0.5 µg pMHC tetramer was added directly to each sample without washing or pre-chilling, followed by incubation for 30 min on ice and in the dark. Finally, cells were washed and analyzed by flow cytometry (SONY, SA-3800). Data were analyzed by FlowJo software.

Artificial feeder cells facilitate the ex vivo expansion of TILs from pancreatic tumor fragments.
From the end of March 2020, resected primary pancreatic tumors from 70 patients who underwent the informed consent procedure were collected. Tumor tissues were minced into 1-3 mm fragments and put into TIL culture as described in the method section. Our initial trials failed to generate TIL culture when fragments were plated in TIL media with highdose IL-2 (data not shown), contrary to the typical outcome of melanoma fragment culture [7,13]. The low success rate of TIL generation could be explained by the relatively low frequency of infiltrating lymphocytes found in pancreatic tumor tissues [9]. Friedman et al. used K562-based artificial APC engineered for costimulatory enhancement (ECCE) as feeder cells to accelerate TIL expansion and greatly improved the generation of TILs from single-cell melanoma suspensions with a low frequency of lymphocytes [10]. We recently developed a new K562-based artificial APC engineered to express membrane-bound IL-21 (mbIL-21) and 4-1BBL or CD70 (Fig. 1a). Unlike 4-1BBL, it was shown that APC, which expresses CD70, mediated costimulation and enhanced antigen-specific CD8 T-cells responses without significant bystander activation [22]. Thus, we examined the ability of three different APCs as feeder cells to strengthen the expansion of TILs from pancreatic cancer.
The visible appearance of viable lymphocytes in each well was a positive TIL culture, as shown ( Figure S1). Table 1 shows TIL cultures generated from 3 cryopreserved tumor fragments and one freshly processed tumor fragment from four patients. We found that 10 of 21 wells (47.6%) produced positive TIL cultures in media containing multiple cytokines (IL-7, IL-15, IL-21) and IL-2.  (Fig. 1b). Adding feeder cells into TIL cultures accelerated TIL growth; many more TILs were generated when ECCE feeder cells were used. After tumor fragments were put into cytokine-only culture for 10-14 days, TIL cultures either did not grow or produced less than 1 million cells per well.
On the other hand, most wells produced positive cultures and yielded more than one million cells per well when we added feeder cells to the TIL culture (Fig. 1b). Feeder cells expressing CD70 generated fewer cells than the other two feeder cells expressing 4-1BBL. Vigorous TIL growth and disappearance of feeder cells were only observed when ECCE cells were used as the feeder cells ( Figure S1). Our results indicated that the overall expansion of TILs was greatly enhanced by feeder cells expressing 4-1BBL, whereas CD70 co-stimulation appeared less effective. The dramatic effect of mbIL-21 on the ex vivo growth of NK cells from peripheral blood (pbNK) or cord blood (cbNK) was not seen with TILs from pancreatic patients [16].

Feeder cells increased the expansion of CD8 + T cells and NK cells.
The rapid disappearance of ECCE feeder cells prompted us to examine the ability to expand NK cells that efficiently kill K562 feeder cells. Flow cytometry analysis of TILs after 10-14 days of culture showed a much higher percentage of CD3 − CD56 + cells when TILs were generated with feeder cells than TILs generated by cytokine alone (Fig. 1c and d).
Among the three groups with feeder cells, co-culture with ECCE cells caused a higher percentage of NK cells (average of 50%), consistent with the observation that a rapid expansion of NK cells led to the elimination of feeder cells ( Figures S2A and S2B). Cytokine alone culture yielded the lowest percentage of NK cells (< 20%), and the two other groups with feeder cells other than ECCE produced an average of 30% of NK cells. CD8 + (CD3 + CD4 − ) T cells were found to be the dominant T-cell subset (> 80%) when feeder cells with 4-1BBL were used. Such a skewing effect was not seen with CD70 as the costimulatory molecule. These results showed a significant difference in the composition of TIL culture when 4-1BBL and CD70 were used as the co-stimulation.

Success in TIL generation from fragments of tumor or para-tumor tissues regardless of patient clinical staging or neoadjuvant treatment
Neoadjuvant chemotherapy or radiation therapy is often used for patients whose tumors were not resettable when they were first diagnosed. Seven patients in our cohort received neoadjuvant therapy (gemcitabine + nabpaclitaxel) before tumor resection (Supplementary data  Table S1). In addition, two patients also received radiation therapy besides gemcitabine and nab-paclitaxel. The other patients were chemotherapy-naive. When we compared the  Cryopreserved-NJ013 1/6(16.7%) 6/6(100%) 3/6(50%) 6/6(100%) Cryopreserved-NJ032 6/6(100%) 6/6(100%) 5/6(83.3%) 6/6(100%) Cryopreserved-NJ025 2/3(66.7%) 3/3(100%) 3/3(100%) 3/3(100%) Fresh NJ043 1/6(16.7%) 3/6(50%) 0/6(0%) success rate of TIL generation from tumor fragments from chemo-naive or neoadjuvant chemo-treated patients, we could not find any significant difference between these two groups (Fig. 2a). Therefore, neoadjuvant chemotherapy did not influence the ability to generate TIL cultures from peritumoral tissues. Most TIL studies have been limited to the tumor tissues; however, a few studies also examine lymphocyte infiltration in peritumoral tissues. Strizova Z et al. performed immune cell profiling of 62 tissue samples from 16 esophageal cancer patients, including tumor tissue, peritumoral tissue, healthy esophageal tissue, and adjacent lymph nodes [23]. Unexpectedly, patients who responded poorly, or not at all, after neoadjuvant chemotherapy showed an increased proportion of peritumoral CD8 + T cells. In our cohort of patients, the TIL cultures were less successful for peritumoral tissues than tumor tissues. Except for three patients, we could generate TIL culture from the tumor tissues of all patients (n = 64). Our current TIL protocol was robust enough to successfully grow at least one TIL culture from more than 90% of patients, whether they received neoadjuvant chemotherapy or not (Fig. 2a). At the same time, only 50% of cultures of peritumoral tissues gave rise to successful TIL growth (Fig. 2b). On average, 50% of wells can yield sufficient T cells for REP procedure regardless of clinical stages (Fig. 2c). We did not observe a similar TIL generation rate from patients irrespective of the levels of CA19-9 (Fig. 2d).

Rare CD8 T cells in TILs were stained by HLA-A*11:01 MHC tetramer complexed with G12V KRAS peptide but disappeared rapidly after ex vivo culture.
Because most PDAC tumors harbor the KRAS driver mutation, we thought to generate T cells specific for this shared neo-antigen. The frequency of different KRAS mutations in our cohort of patients was close to what was reported except for the wild-type KRAS frequency (23.44%) [24]. The specific hotspot KRAS mutations are dominated by G12D (42.19%) or G12V (25%), followed by G12R (6.25%) and Q61H (3.13%). Not surprisingly, we could generate TILs with a similar success rate regardless of KRAS genotypes (Figs. 3a and 3b). In addition, published data indicated that T cells recognize mKRAS presented at a shallow frequency in tumors or memory T-cell population of PBMC from cancer patients [20,[25][26][27]. Like these studies, we identified one TIL culture from patient #16 that was reactive to KRAS G12V peptides (Fig. 3c). We successfully generated 12 TIL cultures from patient #16 primary tumor; reactivity to KRAS G12V was measured by staining with HLA-A11:01 tetramer complexes loaded with KRAS G12V 9-mer or 10-mer peptide ( Figure S4). Of 12 cultures, only the #6 culture was stained positive by tetramer loaded with 9-mer KRAS G12V peptide. Among these expanded TILs from fragment culture, approximately 0.33% of CD8 positive T cells were stained positive by tetramer (Fig. 3c).

Fig. 2
Comparison of ability to generate TILs from tumor fragments of 64 pancreatic cancer patients. a Comparison between the "No Chemo" group (n = 57) and the "Chemo Treat" group (n = 7). Patients in the "Chemo Treat" group received neoadjuvant chemotherapy before removing their primary tumors (Table S1). b Comparison between the tumor tissue group and with para-carcinoma tissue group. The paired tumor and para-tumor tissues were collected from 42 patients. c Comparison of TIL generation success rate from patients' tumors with different clinical stages. d Comparison of TIL generation success rate from patients' tumors with various CA19-9 levels. P values were determined by a two-tailed t-test (no pairing). P < 0.0001 **** The staining was specific for 9-mer KRAS G12V peptide; tetramer complexed with 9-mer wild-type KRAS peptide failed to stain the same culture. Interestingly, this reactivity was limited to 9-mer peptide, and tetramer staining with either wild or G12V 10-mer peptide was negative (Fig. 3d). These results indicated that TILs containing T cells were exclusively specific for KRAS 9-mer mutant epitope; they could discern it from wild-type or mutant peptide with extra N-terminal amino acid (10-mer). Unfortunately, the rare population of tetramer-positive cells quickly became undetectable when we put the positive TIL into continued culture for two more days before sorting the cells for single-cell sequencing. [28]. Therefore, we decided to investigate the possibility of using functional assays that measure the up-regulation of cell surface markers, such as the 4-1BB molecule, as another approach to screen antigen-specificity of TIL cultures [18]. To enable selective growth of mutant KRAS-specific T cells, we examined in vitro sensitization (IVS) methods using aAPC engineered to express TMG of The distribution of different types of KRAS mutations in 64 patients with PDAC was examined in this study. c One well of 24-well plate tumor fragment culture contained mKRASspecific CD8 + T cells that could be identified by MHC tetramer staining. In addition, we generated TIL cultures from NJ016 patient tumor fragments (KRAS G12V mutation, HLA-A*11:01). However, only the #6 well was specifically stained by MHC tetramer loaded with mKRAS9-mer G12V peptide but not by MHC tetramer complexed with either wild type or 10-mer peptide. d Flow cytometry dot plot showed the specific tetramer staining of T cells from one fragment culture. However, positively stained cells disappeared shortly after REP mKRAS. To this end, we engineered the ECCE cells to express HLA-A*11:01 and mKRAS TMG (TMG-ECCE). In addition, Panc-1 tumor cells (HLA-A*1101 and G12D KRAS) were also engineered to express either WT KRAS TMG or mutant TMG containing G12V mutation. These artificial APCs could stimulate effector T cells engineered to express positive control TCR (PCV-TCR) that specifically recognize G12V mutation in the context of HLA-A*1101 molecules ( Supplementary Fig. 5). These APCs were subsequently used as APCs for IVS. First, we generated six pre-IVS TIL cultures from the #51 patient tumor specimens described in the method section. Next, pre-IVS TILs were stimulated for ten more days in the IVS culture supplemented with recombinant human IL-2 and IL-21 (Fig. 4a). Finally, TILs after IVS were restimulated with HLA-A*11:01 restricted KRAS wild type or G12V 9-mer or 10-mer peptides to measure the upregulation of cell surface marker 4-1BB on CD8 positive T cells. Unfortunately, we found no significant expansion of mKRAS-specific T cells (Fig. 4b-e). Instead, a preferential accumulation of T cells reactive to ECCE cells was observed (data not shown). IVS with an allogeneic PANC-1 cell line with matched HLA-A*11:01 and engineered to express KRAS TMG has the same outcome (data not shown).

Design of Experiment (DoE) to optimize key IVS parameters to expand antigen-specific T cells selectively.
Previously, Rosenberg and his colleague primarily used PBMC-derived autologous dendritic cells as the APC for their IVS culture [18,20,25,26]. We used autologous PBMC engineered to express TMG to expand mKRAS-specific T cells as an alternative approach. Transduced T cells could be developed with rapid expansion protocol (REP) to a relatively large number and repetitively. In our preliminary experiments, with the help of the DoE method, we optimized several vital parameters to enable TMG-T cells to efficiently expand spiked T cells transduced to express mTCR that recognizes G12V in the context of HLA-A*11:01 (Fig. 5a). Spiked mTCR-positive cells mimic a rare population of mKRAS-specific T cells in TIL. After IVS and expansion in a medium supplemented with IL-2 and IL-21 for 12 days, we determined the total cell yield and the percentage of T cells that express mTCR by flow cytometry analysis (Fig. 5b). The IVS optimization experiment via the DoE consists of nine groups with two repetitions (Figs. 5c and d). We found that the human AB serum promoted the generation of specific T cells and total T-cell proliferation. As expected, allogeneic PBMC augmented the T-cell proliferation but had no impact on the expansion of mKRAS-specific T cells; however, TMG-T APC could enhance the growth of mKRAS-specific T cells (Fig. 5e and f). The optimized conditions are either condition eight or condition 9. In condition 8, 0.2 million expanded TILs stimulated with 2 million irradiated TMG-T cells in the medium supplemented with 10% huAB and 0.2 million irradiated PBMC. Under this condition, 94.4% or 52.3% of expanded T cells on day 12 were mTCR-positive, and the expansion of TILs was 6.67-fold or 7.83-fold for experiment 1 or experiment 2, respectively ( Fig. 5e and f). Condition 9 used the same parameters as condition 8, except that tenfold higher (2 million) irradiated PBMC was used as the bystander feeder cells. Under this condition, 83.4% or 48.9% of expanded T cells on day 12 were mTCR-positive, and the expansion of TILs was 10.5-fold or 26.9-fold for experiment 1 or experiment 2, respectively ( Fig. 5c and d).
Noticeably, the antigen-driven proliferation of T cells transduced with PCV mTCR was ineffective when fewer TMG-T cells were used in experiment 1 and to a minor degree in experiment 2 despite the excellent expansion of the total T-cell population (condition 7). The allogeneic PBMC feeder and human AB serum promoted the proliferation of whole responder T cells. Still, TMG-T cells were critical for the antigen-driven expansion of mTCR-positive T cells. Thus, condition 8 drives the proliferation of antigen-specific T cells from TILs, whereas condition 9 increases the total number of T cells.

IVS with TMG-T selectively expands T cells reactive to G12V KRAS neo-antigen
Next, we used the optimized condition (8) to expand mKRASspecific T cells from TIL culture of patient #51 (Fig. 6a). We found that 8% of CD8 + T cells in the expanded TILs up-regulated 4-1BB surface expression when we stimulated them with a 10-mer mKRAS peptide (Fig. 6b); they were not reactive to a 9-mer peptide containing the same mutation or 10-mer peptide of wild KRAS (Fig. 6c). The background 4-1BB surface expression was 1.8% of CD8 + TILs; the 4-1BB functional assay would likely have a significantly higher background since in vitro expanded T cells would likely exhibit a low level of antigen-independent T-cell activation (Fig. 6b). MHC tetramer staining had a much lower background than functional assay-based 4-1BB expression. No or low background staining was observed with 9-mer MHC tetramer ( Figures S4); a clearly defined cloud (2.64%) of CD8 + TILs was positively stained when 10-mer mutant KRAS/A11 tetramer was used (Fig. 6c). The staining was highly specific because no staining was found when we used tetramers complexed with 9-mer and 10-mer wt KRAS peptide (not shown) or 9-mer mKRAS peptide (Fig. 6c). The results were reproducible; expanded TILs were highly specific and functional when TMG-T cells were used as the APC (Supplementary Fig. 6). After IVS with TMG-T, over 10% expanded CD8 + TILs were induced to express 4-1BB when restimulated with mutant peptide or tumor cells that expressed proper HLA and KRAS mutation. We have cloned the TCRs for these TIL cultures and found one dominant TCR that was a specific G12 KRAS mutation. The description of this TCR and functional characterization will be reported separately. Therefore, we could successfully use engineered PBMC to express TMG (TMG-T) as autologous APC to expand rare neo-antigen-specific T cells in vitro. TMG-T cells are a reliable source of APC when autologous DC is unavailable.

Discussion
Adjuvant chemotherapy after surgical resection remains the standard of care in pancreatic cancer. Neoadjuvant chemotherapy (radiotherapy) benefits so-called borderline resectable and locally advanced tumors, representing the standard of care resulting in high secondary resection rates. Increasing evidence suggests neoadjuvant chemotherapy has a beneficial effect on the overall survival in CD8 + TILs unregulated surface 4-1BBL expression upon re-stimulation with mutant G12V 10-mer peptide. c IVS expanded TILs were stained with MHC tetramer complexed 10-mer G12V peptide but not a 9-mer peptide. Results showed that approximately 2.6% of CD8 + T cells were stained by MHC tetramer (HLA-A*11:01 complexed with 10-mer G12V mKRAS peptide) resectable PDAC compared to upfront resection. Unfortunately, not all patients respond to neoadjuvant FOL-FIRINOX chemotherapy [29]. TIL culture could reflect the immune response in the PDAC tumor microenvironment, and a combination of neoadjuvant chemotherapy and immunotherapy might benefit PDAC patients. Neoadjuvant chemotherapy could lead to a more significant infiltration of tumor-reactive TIL and a greater success rate of TIL production. However, we did not observe any significant impact of neoadjuvant therapy on TIL generation in a small cohort of patients. We reasoned that most T cells in TIL culture are bystander cells that are not tumor-antigen specific. The bystander T cells in tumors are less likely influenced by neoadjuvant chemotherapy [30][31][32].
The primary goal of the current study is to establish a reliable method for the generation of TILs. With K562based artificial APCs that express 4-1BBL as the bystander co-stimulators, we could successfully develop at least one TIL culture from tumor fragments from most patients with PADC and preferential proliferation of NK cells CD8 + TILs. The preferential expansion of NK cells in the presence of 4-1BBL expression aAPC is consistent with the notion that 4-1BB is the critical costimulatory signal for NK cells [33,34]. Most engineered feeder cells for NK cell expansion require a high-level expression of 4-1BBL and mbIL-21 [35][36][37][38][39]. In support of previous publications [10,22,[40][41][42], this study further demonstrated that TIL production could significantly improve when 4-1BBL expressing aAPC was used. The TIL production was also enhanced when the 4-1BB agonist antibody was added to the TIL culture [9,43]. However, a recent report also showed that co-stimulation either as 4-1BBL expressing aAPC or with a 4-1BB agonist antibody resulted in antigen-independent bystander T-cell activation and NK cell proliferation [44] at the expense of antigen-specific T cells. Multiple approaches are available to boost the expansion of TILs from tumor fragments; however, the possible loss of tumor-specific T cells during in vitro culture remains a significant concern.
The second goal was to expand TILs cultures reactive to the mutant KRAS selectively. We examined two different IVS methods to expand mKRAS-specific T cells. Publications from Dr. Steven A. Rosenberg's group at NCI documented that, on average, tumors from gastrointestinal cancers harbor 100 non-synonymous mutations. Only 1.6% of somatic mutations are recognized by TILs cultured from 62 of 75 patients [26]. From 124 neoantigen-reactive T-cell populations identified in the 75 patients we screened, 99% of the neoantigen determinants appeared unique and not shared between any two patients. However, rare exceptions were also documented. For example, CD8 + T cells from two colorectal patients recognized KRAS G12D in the context of HLA-C*08:02 [1,18]. The rarity of mKRAS-reactive T cells and the low success rate in identifying authentic mKRAS-specific T cells in expanded TIL cultures raised the question of whether these T cells were lost during the TIL expansion phase [26]. Pankhurst and her colleagues demonstrated that TIL reactivity against neo-antigens changed dramatically during the in vitro culture. Neoantigen-reactive T cells gradually decreased to a very low level during the TIL culture and became undetectable after rapid expansion with feeder cells. In their study, the change of TCR composition was unavailable due to a lack of starting materials for TCR deep sequencing. Recently, Poschke and colleagues used TCR deep sequencing to examine the change of TCR composition systemically and in vitro growth capacity of T-cell clones in TILs from primary resectable pancreatic ductal adenocarcinoma [45]. They found a loss of tumor reactivities due to the drastic changes in TCR repertoire during in vitro TIL expansion, particularly the loss of most tumor-dominant T-cell clones and the outgrowth of bystander T-cell clones not prominently present in the initial TIL isolate. Only IVS with autologous T cells expressing tandem minigenes (TMG-T) cells in our current study allowed us to selectively expand neo-antigen-specific T cells from expanded TILs. This protocol will set the foundation for TIL therapy or identifying TCRs specific for shared neoantigens such as mKRAS or patient-specific neo-antigens unique to the individual patient. The frequency of mKRAS reactive T cells (0.0025%) in patients' PB is very low, but recent publications have successfully used the DC-based IVS method to expand mKRAS-specific T cells from the health donors [46,47] with a high success rate.
Currently, TIL therapy has experienced a renaissance due to the encouraging outcomes from multiple phase I/II trials in combination with an anti-PD-1 antibody treatment. However, TIL therapy, in general, and pancreatic cancer in particular, is challenging to implement. Sufficient tumors with abundant TIL are challenging to obtain for pancreatic patients. The manufacturing process for TILs is tedious and nonreliable. The length production process often led to T cells being exhausted and quickly losing their ability to proliferate and persist after adoptive transfer. Because the antigen specificity is not known and reliable product release criteria are difficult to obtain, the regulatory approval for TIL products is often significantly delayed. Many limitations of TIL therapy could be overcome by TCR-based treatment. A collection of TCRs that recognize multiple mKRAS in the context of various common HLA molecules will finally open the gate for the TCR gene-modified adoptive T-cell therapy [25]. The most recent work from Levin and Rosenberg also developed a new TIL IVS method, which combined a TIL screening with coculture of the reactive T cells with dendritic cells (DC) loaded with peptides or transfected with mRNA encoding RAS-mutated antigens. The TIL IVS method identified 12 more mKRAS-specific TCRs and the 9 TCRs identified by TIL screening alone 48 .
Our group has cloned mKRAS-specific TCRs from #051 patient, and detailed characterization and clinical application will be presented in a different report. These anti-RASmutated TCRs can eventually be used as an "off the shelf" treatment for 90% of patients with PDAC whose tumor bears RAS mutations.