T Cells Expanded from PD-1⁺ Peripheral Blood Lymphocytes Share More Clones with Paired Tumor-Infiltrating Lymphocytes

Tumor-infiltrating lymphocytes (TIL) are polyclonal populations enriched for T cells targeting known and unknown tumor-specific antigens, including shared tumor-associated antigens and private tumor neoantigens (1). Clinical benefit from infusion of in vitro–expanded TILs via adoptive cell therapy (ACT) has been observed in patients with malignant melanoma, and other cancer types (2–4). However, to prepare TILs, fresh tumor materials are required, and even when such materials are available, this feat is difficult to achieve for many tumor types. Thus, alternative methods are needed to prepare autologous tumor-reactive T cells when TILs are not available.

Rosenberg and colleagues demonstrated that neoantigen-specific T cells could be isolated from peripheral blood lymphocytes (PBL), especially T cells expressing PD-1, of patients with tumor, providing a robust and scalable strategy to develop highly personalized T-cell products for patients with tumor (5–7). However, these neoantigen-specific T cells' frequency is so low that only through large-scale in vitro expansion can sufficient number of cells for clinical application be obtained (5, 6). Therefore, further investigations on establishing an efficient protocol to isolate and expand PD-1⁺ PBLs are highly warranted. In addition to this, Gros and colleagues showed that there was a relatively high degree of overlap between T-cell receptor (TCR)-β repertoires of the tumor-infiltrating and circulating CD8⁺PD-1⁺ subsets in the absence of in vitro expansion (6). However, it is still unknown whether the TCRβ repertoire of PD-1⁺ PBLs will change dramatically after long-term in vitro expansion. Thus, it is important to address whether T cells expanded from PD-1⁺ PBLs maintain a similar repertoire like paired TILs.

Appropriate stimulation of the immune system with immune checkpoint inhibitors, such as an anti-PD-1 antibody, may cure patients with cancer (8, 9). However, only a minority of patients achieve long-term, durable responses, while resistance is the unfortunate experience for most patients. Thus, there is an urgent need to develop effective treatments when anti-PD-1 therapy fails, especially in patients without driver gene mutations (10, 11). A previous study showed that T cells isolated from patients with checkpoint inhibitor–resistant malignant melanoma were functional and could mediate tumor regression (12). More recently, Creelan and colleagues reported that durable complete responses (CR) to ACT using TILs were observed in patients with non–small cell lung cancer (NSCLC) who progressed following therapy with nivolumab (13). However, until now, there has been no clinical evidence supporting the idea that T cells expanded from PD-1⁺ PBLs can be used to treat anti-PD-1–resistant patients. Thus, further investigations on the outcome of patients treated with T cells expanded from PD-1⁺ PBLs following disease progression during or after anti-PD-1 therapy are highly warranted.

In this study, we have begun addressing these issues. We established a highly efficient protocol to prepare polyclonal T cells from PD-1⁺ PBLs and characterized these cells' phenotypes and functions. Using deep TCRβ sequencing, we determined the relationship between the TCRβ repertoire of the TILs and T cells expanded from paired PD-1⁺ PBLs. Finally, we treated 4 patients, who developed acquired resistance to anti-PD-1, with T cells expanded from PD-1⁺ PBLs combined with anti-PD-1 antibody.

Tumor tissue and peripheral blood (PB) were collected from patients with informed written consent and following the guidelines of the Ethical Committee of the Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital (Zhengzhou, Henan, P.R. China). The study protocol was approved by the Ethical Committee at the Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital (Zhengzhou, Henan, P.R. China). Additional information on patients can be found in Supplementary Table S1.

The following antibodies and reagents were used: biotin-anti-human IgG4 (9200-08, SouthernBiotech), anti-human IFNγ (17731982, eBioscience), anti-human CD3 (300306, BioLegend), Mouse anti-human CD3 (300330, BioLegend), anti-human CD8 (344710, BioLegend), anti-human CD8 (300926, BioLegend), anti-human CD4 (317414, BioLegend), anti-human CD4 (317428, BioLegend), anti-human CD62L (304810, BioLegend), anti-human CD45RA (304106, BioLegend), anti-human CD45 (368504, BioLegend), anti-human PD-1 (329908, BioLegend), HLA-A*02:01 NY-ESO-1 Tetramer-SLLMWITQC (TB-M011-1, MBL), HLA-A*02:01 Mart-1 Tetramer-ELAGIGILTV (TB-0009-1, MBL), OKT3 (T210, Takara), RetroNectin (T202, Takara), human IL-2 (Quanqi), X-VIVO 15 Serum-free Medium (Lonza), CliniMACS Anti-Biotin GMP Microbeads (170076709, Miltenyi Biotec), Ficoll (MD Pacific), and 7-AAD (420404, BioLegend).

PB from patients who had not received anti-PD-1 therapy was first incubated with anti-PD-1 (nivolumab; 10μg/mL) for 1 hour at 4°C to ensure all PD-1⁺ cells bound with the anti-PD-1 antibody. They were then subjected to isolation of PB mononuclear cells (PBMC) by Ficoll density gradient centrifugation. PB from patients who had received anti-PD-1 therapy was directly subjected to PBMC isolation. Then, freshly isolated PBMCs were incubated with biotin-anti-human IgG4 (10 μg/mL) for 10 minutes at room temperature and subsequently incubated with anti-biotin microbeads (20 μL for 10⁷ PBMCs) for 15 minutes at 4°C before transferring into MS or LS column for cell separation. The magnetic separation procedure was repeated once to increase purity. The purity was determined by staining with PE-anti-human IgG4. Cell debris and dead cells were excluded from the analysis on the basis of scatter signals and 7-AAD fluorescence.

Cell culture flasks were precoated with RetroNectin (6 μg/mL) and anti-human CD3 antibody (1.5 μg/mL) for 24 hours at 4°C. An equal number of sorted PD-1⁺ and PD-1⁻ PBLs was suspended with X-VIVO 15 serum-free medium supplemented with IL2 (1,000 IU) and autologous plasma (2%), and then seeded in the precoated flasks. Cells were transferred into a new flask on day 4, and the fresh medium was regularly changed every other day. The number of viable cells was counted with trypan blue staining.

PB samples were collected in preparation tubes and PBMCs were isolated. Single-cell suspensions were obtained, and after lysis of erythrocytes using an ammonium-chloride-potassium buffer, PBMCs were immediately stained. Anti-PD-1 (329908, BioLegend) was used to stain PD-1 on PBLs from patients who had not received anti-PD-1 therapy (pre-samples). Anti-human IgG4-PE staining (9200-08, SouthernBiotech) was performed as described previously (14) to detect PD-1–expressing cells in patients who had received anti-PD-1 therapy (post-samples). PD-1–expressing cells from pre- and post-treatment samples were gated on the basis of their Fluorescence Minus One Control (FMO). Intracellular staining for Ki-67 was performed using Foxp3 Fixation Kit (00-5523-00, eBioscience). Stained cells were analyzed on BD FACSCantoII and analyzed using FlowJo Software (FlowJo, LLC.)

Functional assays were performed on T cells expanded from the PD-1⁺ and PD-1⁻ PBLs from 19 patients whose fresh tumor materials were available as described previously, with minor modifications (15). Resected tumors were mechanically disrupted and digested in RPMI1640 Medium (A1049101, Gibco) supplemented with collagenase I (C0130, Sigma-Aldrich), collagenase II (17101015, Gibco), collagenase IV (C5138, Sigma-Aldrich), and hyaluronidase Type V (H6254, Sigma-Aldrich) for 2–3 hours at 37°C to prepare autologous tumor cells. The resultant tumor cell suspensions were cryopreserved in liquid nitrogen. After in vitro expansion of the PD-1⁺ and PD-1⁻ PBLs (days 12–14), the tumor reactivity of expanded T cells was determined by incubating 0.5–1 × 10⁶ autologous tumor cells with an equal number of expanded T cells in the presence of Brefeldin A (420601, BioLegend) for 6 hours at 37°C. T cells alone served as negative controls. Cells were stained with anti-CD3 (300306, BioLegend), anti-CD4 (317410, BioLegend), and anti-CD8 (344710, BioLegend) before fixation with a Fixation Buffer (420801, BioLegend). Intracellular accumulation of IFNγ was detected by flow cytometry (BD FACSCantoII) using the Permeabilization Wash Buffer (421002, BioLegend) and anti-IFNγ (17731982, eBioscience). The proportion of tumor-reactive T cells was calculated by IFNγ_{(T cells + tumor cells)}% − IFNγ_{(T cells only)}% − IFNγ_(isotype)%.

Tetramer staining was performed according to the manufacturer's protocol. A measure of 10 μL of tetramer (for 0.5–1 × 10⁶ PBMCs or expanded T cells) was added, followed by incubation at room temperature for 30 minutes. Then, 5 μL of a mixture containing anti-CD8 (300926, BioLegend) and anti-CD3 (300329, BioLegend) was added and incubated for another 20 minutes at 4°C. Cells were washed once with PBS and resuspended with PBS supplemented with 0.5% formaldehyde and stored at 4°C for 1 hour before analysis by flow cytometry (BD FACSCantoII). Tetramer⁺ cells were gated on the basis of their FMO. Data analysis was performed using FlowJo software.

RNA was extracted from paired unexpanded TILs and T cells expanded from the PD-1⁺ and PD-1⁻ PBLs using the RNAiso Plus Reagent (9109, Takara). Samples were analyzed by high-throughput sequencing of TCRβ genes using the ImmuHub TCR Profiling System at the deep level (ImmuQuad Biotech). A 5′RACE unbiased amplification protocol was used, and sequencing was performed on an Illumina HiSeq X10 System with PE150 Mode (Illumina). A postsequencing algorithm was applied to raw sequencing data for PCR and sequencing error correction and V, D, J, and C gene segments mapping with the international ImMunoGeneTics information system (http://www.imgt.org). The resulting nucleotide and amino acid sequences of CDR3 of TCRβ were determined, and those with out-of-frame and stop codon sequences were removed from the identified TCRβ repertoire. We further defined each TCRβ clonotype's amounts by adding numbers of TCRβ clones sharing the same amino acid sequence of CDR3. Shared TIL clonotypes were defined as clones found in both tumor and T cells expanded from paired PBLs and nonshared TIL clonotypes as clones found in T cells expanded from PBLs but not in paired TILs. Shared and nonshared TIL clones were subsequently analyzed for their presence and frequency in the T cells expanded from the PD-1⁺ and PD-1⁻ PBLs. The accumulated frequency of TIL clones was defined as the sum (or total) frequency of shared T-cell clones, which could be found in paired unexpanded TILs. Analyses were performed using 8 patients who had sufficient TILs for the assay.

The observational study was conducted at the Department of Immunotherapy, Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital (Zhengzhou, Henan, P.R. China). The study protocol was approved by the Ethical Committee at the Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital. Four patients who failed multiline therapy and developed acquired resistance to anti-PD-1 were recruited (described in detail in Table 1; Supplementary Figs. S1–S4). All 4 patients signed the informed written consent form. The treatment consisted of anti-PD-1 therapy and the infusion of T cells expanded from the PD-1⁺ PBLs. Anti-PD-1 antibodies were administered every 21 days. PB collection was conducted 4 days after anti-PD-1 therapy and cell infusion was given 18 days after anti-PD-1 therapy. The cell infusion was stopped after a total of four cycles, and anti-PD-1 therapy continued every 21 days until disease progression or unacceptable toxicity (cell numbers and the type of anti-PD-1 can be found in Table 1). Radiographic responses were assessed by the investigator according to RECIST v1.1.

All the data are presented as mean ± SEM. Experiments were performed without duplicates, unless otherwise specified. All experiments were repeated at least twice. A paired Student t test (two-tailed) was used for comparison. Statistical analysis was conducted using Prism 8 (GraphPad Software Inc.).

Patients were enrolled on a clinical protocol approved by the Ethical Committee at the Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital (Zhengzhou, Henan, P.R. China) and signed informed written consent.

Considering that an increasing number of patients with cancer are receiving anti-PD-1 therapy, an anti-human IgG4 antibody was used to isolate anti-PD-1–bound PD-1⁺ PBLs (Fig. 1A). Figure 1B shows a representative graph of sorted PD-1⁺ PBLs. The compositions of the sorted PD-1⁺ and PD-1⁻ fractions are shown in Supplementary Fig. S5. The purity of PD-1⁺CD3⁺ T cells obtained from this method was approximately 85% (median, 82.3%). Subsequently, sorted PD-1⁺ PBLs were subjected to in vitro expansion. The proliferation curves (fold change) of sorted PD-1⁺ and PD-1⁻ PBLs are shown in Fig. 1C. The number of PD-1⁺ PBLs increased exponentially during weeks 2–3. Meanwhile, we noticed that the proliferation rate of PD-1⁺ PBLs was significantly slower than PD-1⁻ PBLs, and this result is consistent with a previous study (16). Statistical results from 21 cases indicate that we could obtain 4.72 ± 0.25 × 10⁹ T cells expanded from PD-1⁺ PBLs in 2 weeks using 50 mL PB as starting material (Fig. 1D). Next, we performed flow cytometry to determine the composition of T cells expanded from PD-1⁺ PBLs. We found the majority were CD3⁺ T cells (91.63% ± 0.83%), while the percentage of CD3⁺CD8⁺ T, CD3⁺CD4⁺ T, CD3⁺CD8⁻CD4⁻ T, and CD3⁺CD56⁺ T cells was 73.30% ± 2.16%, 19.51% ± 1.92%, 3.20% ± 0.32%, and 2.94% ± 0.37%, respectively (Fig. 1E). Furthermore, we stained T cells with antibodies targeting CD45RA and CD62L. The results showed that 89.25% ± 1.16% of CD8⁺ T cells and 41.74% ± 4.56% of CD4⁺ T cells exhibited a CD45RA⁺CD62L⁺ phenotype and 6.03% ± 1.26% of CD8⁺ T cells and 48.44% ± 3.82% of CD4⁺ T cells exhibited a CD45RA⁻CD62L⁺ phenotype (Fig. 1F; Supplementary Fig. S6).

Previous studies suggested that anti-PD-1 therapy can induce proliferative responses in peripheral CD8⁺PD-1⁺ T cells (14, 17). Therefore, we determined whether anti-PD-1 therapy impacts in vitro expansion rate of PD-1⁺ PBLs. As illustrated in Fig. 2A–C, anti-PD-1 therapy promoted an increase in not only the frequency of PD-1⁺ PBLs (Fig. 2B), but also the proportion of proliferative PD-1⁺ PBLs (Ki-67⁺PD-1⁺%; Fig. 2C) in both CD8⁺ and CD4⁺ subsets, suggesting that anti-PD-1 therapy enhances in vivo proliferation of PD-1⁺ PBLs, which are similar to previous results (17). Furthermore, we isolated and expanded PD-1⁺ PBLs from the same patient pre- and post-anti-PD-1 therapy. We found that the in vitro expansion rate of PD-1⁺ PBLs from post-anti-PD-1 therapy was significantly faster (∼4 times) than that from pre-anti-PD-1 therapy during the same period (Fig. 2D–F).

To detect tumor-reactive T cells in T cells expanded from PD-1⁺ PBLs, a functional T-cell assay was performed on the basis of the coincubation of T cells with paired tumor cells and the detection of the proportion of IFNγ-expressing T cells (15). The result of a representative experiment is shown in Fig. 3A. CD8⁺ T cells derived from PD-1⁻ PBLs hardly expressed IFNγ when stimulated with the matched tumor cells. In contrast, the frequency of IFNγ⁺ cells in CD8⁺ T cells expanded from PD-1⁺ PBLs was relatively low before stimulation, but increased sharply after stimulation with autologous tumor cells. Meanwhile, similar results were observed in the CD4⁺ subset (Fig. 3B). We further performed this assay with another 18 samples, including five malignant melanomas, six hepatocellular carcinomas (HCC), four renal cell carcinomas (RCC), two NSCLCs, and one esophageal carcinomas. The statistical results of 19 samples are shown in Fig. 3C and D for CD8⁺ subset and Fig. 3E and F for CD4⁺ subset. We observed that the frequency of IFNγ⁺ cells in T cells expanded from PD-1⁺ PBLs was significantly higher than that derived from PD-1⁻ PBLs both for CD8⁺ (5.24% ± 2. 31% vs. 0.65% ± 0.20%; P = 0.048) and CD4⁺ (2.42% ± 0.75% vs. 0.38% ± 0.17%; P = 0.005) subsets. These results suggest that across different cancer types, both CD8⁺ and CD4⁺ T-cell subsets expanded from PD-1⁺ PBLs are enriched for autologous tumor-reactive T cells.

Moreover, the pMHC-I tetramer against NY-ESO-1_157–165 was used to stain T cells from patient MM-6 with >30% tumor cells expressing NY-ESO-1. As shown in Fig. 4A, before in vitro expansion, the frequency of tetramer⁺ cells in PD-1⁺CD8⁺ PBLs was indeed higher than in CD8⁺PD-1⁻ PBLs (Fig. 4A, top). After in vitro expansion, the frequency of tetramer⁺ cells in T cells expanded from PD-1⁻ PBLs was maintained at a similar level, while that of T cells expanded from PD-1⁺ PBLs increased (Fig. 4A, bottom). Furthermore, consistent results were obtained from flow cytometry analysis of samples from patient MM-7 using tetramers against two different antigens (Mart-1_26–35 in Fig. 4B and NY-ESO-1_157–165 in Fig. 4C). These results further support the notion that T cells expanded from PD-1⁺ PBLs are enriched for tumor-reactive T cells.

TILs harbor tumor-reactive clones, most of which target unknown antigens. Thus, it could be interesting to determine the relationship between the TCRβ repertoire of the TILs and T cells expanded from paired PD-1⁺ PBLs. We analyzed the number of amino acid CDR3 clonotypes shared between each pair of TILs (unexpanded) and T cells expanded from PD-1⁺ and PD-1⁻ PBLs. These PD-1⁺ and PD-1⁻ PBLs were derived from 8 patients whose TILs were sufficient for the assay. As shown in Fig. 5A and B, an average of 11.29% (1.32%–29.06%) of tumor-resident clonotypes were found in T cells expanded from paired PD-1⁺ PBLs, and this proportion was significantly higher than T cells expanded from PD-1⁻ PBLs (P = 0.015; n = 8). Furthermore, we calculated the accumulated frequency of shared TIL clones in T cells expanded from PD-1⁺ PBLs. Although there were differences among different individuals, an average of 35.11% (7.23%–78.02%) of T cells expanded from PD-1⁺ PBLs overlapped with matched TILs, which was approximately 3.5-fold higher than T cells expanded from PD-1⁻ PBLs (P = 0.017; n = 8; Fig. 5C and D). Moreover, we observed that more top 50 TIL clones were readily identifiable in T cells expanded from PD-1⁺ PBLs than in T cells expanded from PD-1⁻ PBLs (Fig. 5E). These results suggest that T cells expanded from PD-1⁺ PBLs share more clones with paired TILs.

Considering that T cells expanded from PD-1⁺ PBLs are enriched for clones shared with paired TILs, we wondered whether T cells expanded from PD-1⁺ PBLs can be used as TIL substitutes to treat patients with tumor. Four patients who failed multiline therapy and developed acquired resistance to anti-PD-1 were recruited (information on previous treatments can be found in Supplementary Figs. S1–S4). Their baseline characteristics are shown in Table 1. Schematic of combination therapy with autologous T cells expanded from PD-1⁺ PBLs and anti-PD-1 antibody is illustrated in Fig. 6A. Objective responses (OR) according to RECIST1.1 were observed in 3 of 4 patients. The best OR of individual patients is illustrated in Fig. 6B. Percentage changes in target lesion size compared with baseline for 4 patients are shown in Fig. 6C. CT scan results are shown in Fig. 6D. There were no acute infusion-related toxicities and no autoimmune adverse events. No treatment-related mortality occurred. In summary, combination therapy with autologous T cells expanded from PD-1⁺ PBLs and anti-PD-1 antibody resulted in one CR, two partial responses (PR), and one stable disease (SD) in patients with tumor who developed acquired resistance to anti-PD-1 therapy, suggesting that these cells can be used to treat patients with tumor.

Although PD-1⁺ PBLs are enriched for neoantigen-specific T cells, studies on the relationship between T cells expanded from PD-1⁺ PBLs and paired TILs, which are the main source to prepare autologous tumor-reactive T cells, have been limited (5, 6). Here, we first established a highly efficient protocol to prepare polyclonal T cells from PD-1⁺ PBLs. Then, we demonstrated that these cells were enriched for tumor-reactive T cells. More interestingly, we revealed that T cells expanded from PD-1⁺ PBLs share more clones with paired TILs. Furthermore, we observed ORs in 3 of 4 patients, who failed multiline therapy and developed acquired resistance to anti-PD-1, after combination therapy with autologous T cells expanded from PD-1⁺ PBLs and anti-PD-1 therapy.

The clinical results with ACT using TILs are encouraging, but the availability of fresh tumor materials is one of the challenges limiting the number of patients eligible for treatment (2, 3, 18). Cohen and colleagues demonstrated that neoantigen-specific T cells could be isolated from peripheral lymphocytes, implying that the PB of patients with tumor might be a noninvasive choice to prepare autogenous tumor-reactive T cells (7). Moreover, Gros and colleagues showed that the expression of PD-1 could guide the identification of neoantigen-specific lymphocytes in PB of patients with malignant melanoma (6) and gastrointestinal cancer (5). However, the low frequency of these neoantigen-specific T cells among the PD-1⁺ subset in PB may limit the therapeutic efficacy (6). Meanwhile, the expansion rate of PD-1⁺ T cells is approximately 10 times slower than their PD-1⁻ counterparts (16). These studies suggest that establishing a highly efficient and rapid method for in vitro expansion of PD-1⁺ PBLs is a prerequisite for their clinical application. Here, we established a protocol from which we could obtain an average of 5 × 10⁹ T cells expanded from PD-1⁺ PBLs in 2 weeks using 50 mL PB as starting material (Fig. 1D). We observed that after in vitro expansion, a portion of T cells derived from PD-1⁺ PBLs (most of CD8⁺ T cells and less than half of CD4⁺ T cells) expressed CD45RA and CD62L, suggesting that these T cells have stronger homing ability (Fig. 1F). Although naïve T cells express CD45RA and CD62L, we are unsure whether these in vitro–expanded CD45RA⁺CD62L⁺ T cells can be classified as “naïve T cells” like these unexpanded T cells in PBLs. Therefore, we would rather consider these in vitro–expanded CD45RA⁺CD62L⁺ T cells express and possess molecular functions of CD45RA and CD62L than call them “naïve T cells.”

Having established this highly efficient and rapid protocol, we investigated whether these T cells expanded from PD-1⁺ PBLs were enriched for tumor-reactive T cells. Using a functional T-cell assay (19), we demonstrated that it might be a universal phenomenon across different cancer types (19 patients, including five malignant melanomas, seven HCCs, four RCCs, two NSCLCs, and one esophageal carcinoma) that T cells expanded from PD-1⁺ PBLs were enriched for more autologous tumor-reactive T cells both in CD8⁺ and CD4⁺ T-cell subsets (Fig. 3). Using tetramers targeting NY-ESO-1_157–165 and Mart-1_26–35, we further showed that T cells expanded from PD-1⁺ PBLs using the protocol established in this study harbored more antigen-specific T cells (Fig. 4). These results support and further extend the notion that PD-1 can guide the identification of patient-specific tumor-reactive lymphocytes from PB (5, 6).

Considering that both T cells expanded from PD-1⁺ PBLs and TILs are enriched for tumor-reactive clones, it is important to determine the extent of the overlap between their TCRβ repertoires, although some studies reported low and variable tumor reactivity of the intratumoral TCR repertoires in certain cancer types (15, 20). Although there were differences among different individuals, we found an average of 11.29% (accumulated frequency, 35.11%; n = 8) of TIL clonotypes in T cells expanded from paired PD-1⁺ PBLs (Fig. 5). This experiment's setting is different from the previous study, in which the analysis of TCRβ repertoires of intratumoral and peripheral CD8⁺PD-1⁺ in the absence of in vitro expansion was shown (6). However, the results were similar. This result revealed that T cells expanded from PD-1⁺ PBLs using the protocol established in this study shared more clones with paired TILs and suggested that the TCRβ repertoire of PD-1⁺ PBLs might not change dramatically during the expansion with this protocol, which is consistent with our observation that the cumulative frequency of shared TIL clones in TCRβ repertoire of PD-1⁺ T cells from patient MM-4 was maintained at a similar level before and after in vitro expansion (Supplementary Fig. S7). However, this result should be interpreted cautiously. Poschke and colleagues found that the TIL TCR repertoire changes dramatically during in vitro expansion, leading to a loss of tumor-dominant T-cell clones and the overgrowth of newly emerging T-cell clones (21). These results suggested that even if the same expansion method was used, long-term in vitro expansion would cause the proportions of individual TCRs to become different after expansion, let alone using different expansion protocols. Considering the ex vivo TIL expansion outcome is highly influenced by differences in intrinsic in vitro growth capacity between T-cell clones, we chose unexpanded TILs instead of expanded TILs as the reference for tumor-reactive T cells in this assay.

Compared with the other seven samples included in Fig. 5, we noticed fewer TCR clonotypes (805) had been identified, and T cells underwent a limited clonal expansion in TIL samples from patient MM-3. We tend to think this phenomenon is a true representation of the T cells infiltrating states in this sample. To depict more clearly, the composition of this sample is shown in Supplementary Fig. S8. The highest TIL clone accounted for 2.04%, while the top 10 TIL clones accounted for 6.92% of TIL clones. Considering this patient had malignant melanoma, these numbers are small, although a limited expansion of TIL clones was observed in other tumor types, like pancreatic ductal adenocarcinoma (22). In addition, we cannot exclude the possibility that the tumor's spatial heterogeneity might contribute to this phenomenon (22–24).

Until recently, no standard immunologic treatments existed for patients who have progressed to anti-PD-1 therapy (10, 11). Previous studies have shown that TILs prepared from patients with anti-PD1–resistant malignant melanoma and NSCLC can mediate tumor regression (12, 13). Considering that T cells expanded from PD-1⁺ PBLs harbor more TIL clones, we conducted an observational study to treat patients, who failed multiline therapy and developed acquired resistance to anti-PD-1, with autologous T cells expanded from PD-1⁺ PBLs combined with anti-PD-1 antibody. Among the 4 patients, 3 had OR (including two PR and one CR; Fig. 6). Given this encouraging result, a clinical trial (ClinicalTrials.gov identifier: NCT04268108) is in progress at our center. Meanwhile, we cannot rule out the possibility that it is the rechallenge of anti-PD-1, but not the transfer of T cells expanded from PD-1⁺ PBLs, that mediates tumor regression, although the response rate of anti-PD-1 rechallenge in patients who progressed on the first anti-PD-1 treatment is low (25–27). Randomized controlled clinical trials are needed to elucidate this question.

In summary, our data demonstrate that using the protocol established in this study, we can obtain sufficient polyclonal T cells expanded from PD-1⁺ PBLs for the clinical application in a short time. We further demonstrate that these cells are enriched for more tumor-reactive T cells. More interestingly, these cells share more clones with paired TILs and could be used to treat patients with tumor as TIL substitutes.

No disclosures were reported.

T. Li: Conceptualization, funding acquisition, investigation, methodology, writing–original draft. L. Zhao: Resources, data curation, investigation, writing–review and editing. Y. Yang: Resources, data curation, investigation, writing–review and editing. Y. Wang: Resources, investigation, methodology. Y. Zhang: Resources, data curation. J. Guo: Resources, data curation. G. Chen: Resources, data curation. P. Qin: Resources, data curation. B. Xu: Resources, data curation. B. Ma: Resources, data curation. F. Zhang: Resources, data curation. Y. Shang: Resources, data curation. Q. Li: Resources, data curation. K. Zhang: Resources, data curation. D. Yuan: Data curation, software. C. Feng: Data curation, software, writing–review and editing. Y. Ma: Resources, data curation. Z. Liu: Resources, data curation. Z. Tian: Resources, data curation. H. Li: Writing–review and editing. S. Wang: Writing–review and editing. Q. Gao: Conceptualization, supervision, funding acquisition, methodology, project administration, writing–review and editing.

This work was funded by the National Natural Science Foundation of China (grant no. 81902902 to T. Li), Henan Medical Science and Technique Foundation (grant nos. SBGJ202002027 to T. Li and 201701030 to Q. Gao), and Henan Provincial Scientific and Technological Project (grant no., 182300410344 to Q. Gao). The authors thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.