The patient diagnosed with advanced CDC and multiple bone metastases at Mengchao Hepatobiliary Hospital of Fujian Medical University was enrolled in this study. The CDC biopsies were collected at preimmunotherapy time point and after three treatment cycles of the neoantigen-based immunotherapy. Corresponding blood samples were collected at the same time points. Available CT/MRI scans were also acquired. This is a single-centre, single arm clinical study of an individualized neoantigen peptide vaccine combined with NRTs in the treatment of advanced malignancies retrospectively registered in 17 August 2018 (http://www.chictr.org.cn/showproj.aspx?proj=30138).

Tumor tissues obtained by needle biopsy and peripheral blood samples were used for whole-exome sequencing (WES) and RNA sequencing. In brief, after DNA/RNA extraction and library construction, both enriched exome and transcriptome libraries were sequenced by Annoroad Gene Tech. (Beijing) Co. on Illumina HiSeq X10 platform (paired end, 150 bp).

WES read pairs were aligned to UCSC human reference genome hg19 (GRCh37) using BWA.11 Somatic mutations were detected with Mutect2 and Somaticsniper, and then further confirmed by DeepSNV.12 Indels were detected by Metect2 and Strelka, and manually confirmed by Integrative Genomics Viewer at both DNA and RNA levels. For RNA-seq, the qualified reads were aligned to human genome reference (GRCh37) with GENCODE gene annotation using STAR.13 To quantify the expression levels for each gene, transcripts per kilobase per million mapped reads (TPM) were calculated.

To keep high-quality somatic mutations for following neoantigen identification, all mutations consistently identified by Mutect2 and Somaticsniper were further filtered with following criteria: ≥20× depth in both tumor tissues and peripheral blood samples, variant reads≥5 and variant allele frequency (VAF)≥10% in tumor tissues, and VAF≥1% in blood samples. All the qualified mutations were further confirmed at RNA level by DeepSNV14 that mutations≥20× depth and VAF≥10% in tumor tissue were kept, and the expression level of their corresponding genes should be with TPM≥1. All qualified mutations were annotated by variant effect predictor.15 For each nonsynonymous mutation, all possible peptides containing mutated amino acids with length ranging from 8 to 11 mer were extracted, and their binding affinity to corresponding patient’s human leukocyte antigen (HLA) class I alleles (HLA-A, HLA-B and HLA-C) was predicted by NetMHCpan V.4.0. Meanwhile, the binding affinity of 15 mer peptides containing mutated amino acids to patient’s HLA class II alleles (HLA-DRB1) was also evaluated using NetMHCpanII V.3.1.16 17 Peptides with IC₅₀<500 nM were predicted to be candidate epitopes.

Clonal evolution analysis of preimmunotherapy and postimmunotherapy tumor samples was conducted by SciClone ¹⁸ using all the qualified somatic mutations passing aforementioned criteria.

Patient four-digit HLA class I genotype was assessed by OptiType,19 and the HLA class II alleles were assessed by Seq2HLA20 using RNA-seq data. Additionally, the HLA typing was further confirmed by PCR sequence-based typing.

The therapeutic duration is 180 days for this patient; days 0–60 were the period for neoantigen peptide vaccine preparation. Afterwards, each 30 days were assigned as one treatment cycle. Thirteen personalized neoantigen peptides (27 amino acids in length) were synthesized by the standard solid-phase synthetic peptide chemistry and purified using reverse-phase high-performance liquid chromatography (GenScript USA) with strict quality control (>98% purity, endotoxin concentration was less than 0.01 EU/g and TFA (trifluoroacetic acid) residue was less than 1%). All peptides were divided into three pools containing four to five peptides for each (designated as pools A-C, 0.3 mg/peptide) and mixed with 0.5 mg poly:IC (polyinosinic-polycytidylic acid injection, Guangdong South China Pharmaceutical. Co) as the peptide vaccines. Each pool was administered by subcutaneous injection to one of the limbs (right armpit, left axilla, right groin and left groin).

One hundred millilitres of the patient’s own blood was collected by monocyte blood collection machine, then the PBMCs were isolated and cultured in culture dish for 2 hours to further separate dentritic cells (adherent cell) and T cells (non-adherent cell). Afterwards, the adherent cells were cultured with medium (1% autologous serum+KBM581+1000 U/mL rhGM-CSF (recombinant human granulocyte-macrophage colony-stimulating factor)+500 U/mL interleukin (IL)-4) for 48 hours, and then loaded with 13 neoantigen peptides (25 µg/peptide) accompanied by certain cytokines (tumor necrosis factor alpha, 1000 U/mL; IL-1b, 10 ng/mL; IL-6, 1000 U/mL) in the medium to induce DC maturation overnight. The maturated DCs were cocultured with T cells (0.5–1×10⁶ cells/mL) at a ratio of 1:10 in a culture bag containing 5% autologous plasma+KBM581+100 U/mL IL-2+20 ng/mL IL-7 for 48 hours. Afterwards, the primary NRT cells were collected and stimulated with fresh matured DCs for another two times. The proportion of activated T cells (CD3⁺ 4-1BB⁺) in NRT preparation was assessed by flow cytometry. Then, the T cells were cocultured with OKT3 antibody for amplifying neoantigen-specific T cells to large numbers. On day 14, the prepared NRT cells were washed and resuspended with normal saline and administered to quality control (endotoxin testing ≤5 EU/mL, a negative result for mycoplasma and sterile detection) before cell infusion.

IFN-γ secretion of neoantigen-specific T cells was measured by ELISPOT using Human IFN-γ ELISpot^plus kit (MABTECH) according to the manufacturer’s instruction. Briefly, the DC cells (2×10⁴) were stimulated with neoantigen peptide pool (2 µg in total, 0.15 µg per peptide) or each synthesized neoantigen peptide (2 µg) for 24 hours at 37°C with 5% CO₂. Then, 2×10⁵ PBMCs were pretreated with IL-2 (50 U/mL) for 24 hours and cocultured with the peptide-simulated DC cells (2×10⁴) at 37°C for another 24 hours in a 96-well plate. The plates were washed and subsequently incubated with detection antibody (7-B6-1,1 μg/mL) for 2 hours at room temperature. The plates were washed again and then incubated with streptavidin–horseradish peroxidase (1:1000 dilution) for 1 hour at room temperature. Subsequently, TMB (3, 3', 5, 5'-Tetramethylbenzidine) substrate solution was added to each well and incubated for 15–25 min at room temperature before adding deionized water to stop the reaction. Finally, the spots were imaged and analyzed.

A 73-year-old Asian patient was first admitted to Mengchao Hepatobiliary Hospital of Fujian Medical University due to tenderness pain of spinal and percussive pain of renal. Routine blood and biochemical tests showed no obvious abnormalities. MRI scan revealed a 4.2×3.7 cm nodular in the right kidney, accompanied by multiple bone metastases (L1: 1.4×1.0 cm, L2: 1.8×1.4 cm, L5: 1.5×2.0 cm). As shown in figure 1A, the kidney nodule and bone marrow biopsies were diagnosed as metastatic CDC by a pathologist using H&E examination. Then, the patient subsequently underwent vertebroplasty and sorafenib treatment (400 mg, two times per day) within 2 months. Unfortunately, the patient suffered extensive tumor progression in CDC (MRI scan: 5.7×4.2 cm) and bone metastases (L1: 2.3×2.0 cm, L2: 2.5×2.2 cm, L5: 2.3×1.3 cm).

As poor performance status of the standard first-line treatment and lack of other treatment options, the patient underwent personalized neoantigen immunotherapy based on his own tumor somatic mutation profile. To obtain his tumor-specific neoantigens, a needle biopsy collected from CDC and matched peripheral blood mononuclear cells (PBMCs) were subjected to whole exome sequencing and transcriptomic sequencing. In total, 48 qualified nonsynonymous somatic mutations and seven somatic indels were identified (online supplementary tables S1 and S2), while only 18 of these somatic mutations and 2 of these indels were confirmed at RNA level. To identify reliable neoantigens for clinical preparation of peptide vaccine and NRT cells, we only included 13 somatic mutations with VAF>10% at both DNA and RNA level. Binding affinity prediction of the mutated peptides showed that peptides generated from 12 of the 13 mutations had high affinity (IC50<150 nM) to HLA class I or class II alleles, including refractory mutated genes such as KRAS G12D (table 1). Considering that IC50 of the binding affinity for LIPE A857E peptide with HLA-DRB1*12:02 was 536 nM (near the filtering cut-off value of 500 nM), all the 13 mutations were selected for synthesizing neoantigen peptides. As long peptides may be more immunogenic than short peptides, long peptides (27aa) with the mutated amino acid in the middle were synthesized for the preparation of both peptide vaccine and NRTs.

The patient underwent a total of six treatment cycles of neoantigen-based immunotherapy (figure 1B). For neoantigen peptide vaccine treatment, the vaccine was given on days 1, 5, 8, 15, and 22 of treatment cycle 1 (priming phase) and was given on day 1 of subsequent treatment cycles (boosting phase). Additionally, for the NRT infusion, a total of 5×10⁹ prepared NRT cells were intravenously administered in a 60 min period on day 10 of each treatment cycle, followed by continuous intravenous injection of IL-2 at 3.0 MIU for 4 days. According to the standard image diagnosis, the patient reported stable disease status in tumor burden and significant alleviation of bone pain (cut the painkiller in half) after undergoing neoantigen immunotherapy within 3 months (figure 1C,D). With immunotherapy continuously undergoing, no obvious progression in tumor burden and no obvious side effects have been observed in this patient for another three treatment cycles. During the immunotherapy and follow-up, the patient's ECOG (Eastern Cooperative Oncology Group) performance status score was steadily maintained at level 2. Unfortunately, the patient died after 3 months of finishing neoantigen-based treatments due to complications caused by an accidental fall.

PBMCs were collected from this patient at preimmunotherapy time point, and after three treatment cycles of neoantigen-based immunotherapy (postimmunotherapy) to evaluate the presence of the antigen-specific T cells. ELISPOT assay was performed to assess the immune response following stimulation with neoantigen pools and each neoantigen individually. As shown in figure 2A, the ELISPOT results indicated that the reactivity of PBMCs against total neoantigens were significantly increased in the postimmunotherapy time point when compared with the preimmunotherapy time point. Moreover, based on the results of each individual neoantigen, 12 of the 13 mutated peptides could significantly induce strong immune responses at postimmunotherapy time point, suggesting the generation of potent immune responses against multiple predicted epitopes for this patient (figure 2B,C and online supplementary figure S1A). For 7 of these 12 responsive neoantigens, the T cells could simultaneously respond to mutated peptides and their corresponding wild-type peptides, although the response degree to mutated peptides was higher, suggesting the high degree of cross-reactivity. For another remaining five neoantigen peptides, CCNA1 L426R, LIPE A857E and KRAS G12D showed preferentially immune reactivity to T cells (>10-fold sensitivity) comparing with their corresponding wild-type peptides; the DHX29 V1219E and PAIXP1 Y738C were considered as suspected responses due to the obvious differences between two duplicate wells. In contrary, MTM1 I464S peptide showed no obvious immune reactivity to T cells, while its wild-type peptide showed strong immune reactivity. Moreover, we also analyzed the T-cell activation proportion of NRT preparation by flow cytometry analysis by evaluating CD3⁺/4-1BB⁺ T-cell percentage. As expected, the CD3⁺/4-1BB⁺ T-cell percentage was increased from 1.92% in precultured PBMCs to 7.92% in NRT preparation (figure 2D), suggesting that the NRT cells accounted for about 6% in total infusion of this patient. Here, we only achieved modest enrichment of NRT cells, which might be associated with the enrolled patient’s status, since the enrichment and activity of reactive T cells are heavily relied on the immune cell status; in our study, the patient was relatively old and underwent target drug therapy (sorafenib) before undergoing our neoantigen-based immunotherapy; these issues are well known to impair the functions of the patient’s immune cells.

To investigate the effects of neoantigen-based immunotherapy on immune microenvironment, another CDC biopsy collected at postimmunotherapy time point was subjected to whole-exome sequencing and transcriptomic sequencing. First, we compared the allele frequencies of neoantigens on genomic level between biopsy from preimmunotherapy and postimmunotherapy. The result showed that the mutant allele frequency corresponding to 92% (12/13) of the neoantigens decreased (figure 3A), indicating that the immune system could indeed kill tumor cells carrying these neoantigens. Meanwhile, the mutated allele frequency of MTM1 I464S was upregulated after immunotherapy, which was well consistent with the ELISPOT examination result. However, the allele frequencies for six of eight non-targeted mutations were also decreased in the postimmunotherapy CDC biopsy, while the other two non-targeted mutations were increased (online supplementary table S3). This phenomenon might be attributed to the coexistence of the non-targeted mutations and targeted neoantigens in tumor cells; when NRTs specifically killed tumor cells carrying such neoantigens, the frequency the non-targeted mutations possessed by the same tumor cells also decreased; additionally, the cytokines released by neoantigen-specific T cells might also have non-specific cytotoxicity on other tumor cells. Furthermore, eight new somatic mutations were identified and five of them might be new neoantigen candidates (C2CD5 V874L, IKBKB W291L, LINS L258F, TYK2 R1159C and ZDHHC20 Q339H) in the CDC biopsy after immunotherapy.

The clonal structure dynamics of CDC biopsy from preimmunotherapy and postimmunotherapy showed subtle changes in tumor clonal structure under the pressure of immunotherapy (figure 3B). Interestingly, HLA genes including MHCI (HLA-A, HLA-B and HLA-C) and MHCII (HLA-DRB1), HLA subtypes encoded by which have potential binding affinity to identified neoantigens were all upregulated in CDC biopsy after immunotherapy (online supplementary table S4 and S5). Contrarily, the expression of other HLA subtypes was mostly downregulated in CDC biopsy after immunotherapy. These results suggested that the neoantigen-based immunotherapy might promote the antigen presentation process and immune infiltration for upregulating corresponding HLA subtypes that could potentially bind with neoantigens.

Meanwhile, we also discovered that some classical costimulatory factors showed a certain upregulation in the patient’s tumor tissue after three cycles of neoantigen-based immunotherapy, such as CD70 and 4-1BB, all of which have been reported to be key regulators in the activation of T cells21 (figure 3C). Correspondingly, as commonly reported in immunotherapy, several coinhibitory immune checkpoint molecules also showed a certain upregulation, such as IDO1 and LAG3 (figure 3D), suggesting a negative feedback regulation of T-cell activation. Furthermore, we also obtained the immunophenoscore diagrams of the tumor before and after immunotherapy to show the immune landscape changes.22 However, no obvious differences were observed (figure 3E,F and online supplementary figure S1B).

In summary, our results provide the first evidence to support the feasibility of personalized neoantigen-based immunotherapy for treating CDC with low mutational burden.