In order to explore the molecular mechanism of immunotherapy resistance in OSCC, plasma samples were collected from patients with OSCC who underwent treatment with an anti-PD-1 monoclonal antibody combined with chemotherapy after the fourth treatment cycle between January 2021 and January 2022. The cohort was comprised of 31 patients including 8 females and 23 males. Their ages ranged from 42 to 80 with a median age of 65 years. The inclusion criteria were as follows: i) Age, 18 to 80 years; ii) Eastern Cooperative Oncology Group (ECOG) performance status (PS) score 0 to 1(29); iii) history or pathology confirmed ESCC; iv) chemotherapy regimen consisted of cisplatin plus 5-fluorouracil; and v) no previous tumor-related treatments. The exclusion criteria were: i) esophageal adenocarcinoma and small-cell carcinoma; and ii) incomplete follow-up data. The treatment efficacy was evaluated by the Responsive Evaluation Criteria in Solid Tumours (RECIST) 1.1(30) at the same period. Finally, the plasma samples were divided into the CR/PR group (n=15) and the SD/PD group (n=16). Subsequently, the differentially expressed proteins (DEPs) and differentially expressed metabolites (DEMs) between the two groups, which may be the main contributing factors to immunotherapy resistance, were determined using proteomic (DIA proteomics) and metabolic analysis (including TM-widely targeted metabolomics and widely targeted lipidomics). Furthermore, the DEPs and DEMs were annotated to Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis and the molecular mechanisms were elucidated by conjoint analysis of the DEPs and DEMs. The study work flow is illustrated in Fig. 1.

A total of 31 plasma samples of oesophageal squamous cell carcinoma, including 15 CR/PR and 16 SD/PD who underwent PD-1 monoclonal antibody combined with chemotherapy were collected. The CR, PR, SD, PD was determined using RECIST 1.1. TNM staging was performed according to the sixth Union for International Cancer Control (UICC) TNM classification system (31) due to the large span of diagnosis. All oesophageal samples were collected from patients at Renmin Hospital of Wuhan University (Wuhan, China) between January 2021 and January 2022. Blood was collected after obtaining written informed consent from the participants. All patient consents were provided by the participants themselves or their guardian. The present study was performed in accordance with the Declaration of Helsinki, and the study was approved (approval no. WDRY2020-K212) by the Bioethics Committee of Renmin Hospital of Wuhan University (Wuhan, China). Detailed clinicopathological data and qualified blood samples were available for all participants. The sample details are provided in Table I. All samples were immediately stored at -80˚C until further analysis. The workflow of the present study is presented in Fig. 1.

The steps for sample extraction were as follows: i) The sample was removed from the -80˚C refrigerator and placed on ice until no pieces of ice were observed in the sample (all subsequent operations were required to be performed on ice). ii) After thawing, the sample was shaken on a vortex mixer for 10 sec to ensure thorough mixing, and then 50 µl of the sample was transferred to a centrifuge tube numbered correspondingly. iii) A metabolite extraction agent (300 µl, 20/80 acetonitrile/methanol solution) containing internal standards was added to the centrifuge tube, shaken on the vortex mixer for 3 min, and centrifuged at 1,609.92 x g for 10 min at 4˚C. iv) Following centrifugation, 200 µl of the supernatant was transferred into another correspondingly numbered tube, which was then allowed to stand for 30 min at -20˚C in the refrigerator. v) The aforementioned tube was centrifuged at 1,609.92 x g for 3 min at 4˚C, then 180 µl of supernatant was transferred into the liner tube of the corresponding injection vial for instrumental analysis.

The samples were firstly analysed by a non-targeted metabolite to enlarge the database for widely targeted metabolites. Non-targeted metabolite detection was conducted using ultra performance liquid chromatography (UPLC; ExionLC AD; https://sciex.com.cn/) coupled with quadrupole-time of flight mass spectrometry (TripleTOF 6600; AB SCIEX). The operating conditions of UPLC were as follows: ACQUITY HSS T3 columns (2.1x100 mm, 1.8 µm), 0.1% formic acid/water as mobile phase A, 0.1% formic acid/acetonitrile as mobile phase B, column temperature of 40˚C, flow rate of 0.35 ml/min, and injection volume of 5 µl.

The data acquisition instrumentation system for widely targeted metabolomics primarily comprised a UPLC instrument (ExionLC AD; https://sciex.com.cn/) coupled with a tandem mass spectrometry (MS/MS) instrument (QTRAP^®; https://sciex.com/).

Chromatographic separation conditions were as follows: i) a Waters ACQUITY UPLC HSS T3 C18 column (1.8 µm; 2.1x100 mm) as the chromatography column; ii) ultra-pure water (containing 0.1% formic acid) as phase A and acetonitrile (containing 0.1% formic acid) as phase B; iii) a mobile-phase gradient with water/acetonitrile (V/V) of 95:5 at 0 min, 10:90 at 11.0 min, 10:90 at 12.0 min, 95:5 at 12.1 min, and 95:5 at 14.0 min; and iv) mobile-phase flow rate of 0.35 ml/min, column temperature of 40˚C, and injection volume of 2 µl.

The electrospray ionization (ESI) source temperature was 500˚C, with an ion spray voltage of 5,500 V for the positive mode and -4,500 V for the negative mode. The ion source gas I (GS I) and gas II (GS II) pressures were both 50 psi, and the curtain gas (CUR) pressure was 25 psi. Collision-activated dissociation (CAD) parameters were set to high values. In the triple quadrupole (Qtrap), each ion pair was scanned for detection based on the optimized declustering potential (DP) and collision energy (CE).

The steps for sample extraction were as follows: i) The sample was removed from the -80˚C refrigerator and placed on ice until no pieces of ice were observed in the sample. ii) After thawing, the sample was shaken on a vortex mixer for 10 sec to ensure thorough mixing, and then 50 µl of the sample was transferred to a centrifuge tube numbered accordingly. iii) To the centrifuge tube 1 ml of lipid extraction buffer (methyl tert-butyl ether/methanol=3:1, V/V) was added containing internal standards, followed by shaking the resulting mixture on a vortex mixer for 15 min. iv) To the above mixture 200 µl of water was added, and the new mixture was shaken on a vortex mixer for 1 min, followed by centrifugation at 1,609.92 x g for 10 min at 4˚C. v) Following centrifugation, 200 µl of the supernatant was transferred into a correspondingly numbered centrifuge tube and concentrated until completely dry. vi) To the aforementioned centrifuge tube, 200 µl of mobile phase B was added, and the mixture was shaken on a vortex mixer for 3 min, followed by centrifugation at 1,609.92 x g for 10 min at 4˚C, and then by sampling of the supernatant for LC-MS/MS analysis.

The data acquisition instrumentation system consisted primarily of a UPLC instrument (ExionLC AD) coupled with a tandem mass spectrometry (MS/MS) instrument (QTRAP^®).

Liquid phase conditions were mainly the following: i) A Thermo Accucore^™ C30 column (2.6 µm, 2.1x100 mm) as the chromatography column; ii) acetonitrile/water (60/40, V/V; containing 0.1% formic acid and 10 mmol/l ammonium formate) as mobile phase A, and acetonitrile/isopropanol (10/90, V/V) (containing 0.1% formic acid and 10 mmol/l ammonium formate) as mobile phase B; iii) a mobile-phase gradient with A/B (V/V) of 80:20 at 0 min, 70:30 at 2 min, 40:60 at 4 min, 15:85 at 9 min, 10:90 at 14 min, 5:95 at 15.5 min, 5:95 at 17.3 min, 80:20 at 17.5 min, and 80:20 at 20 min; iv) a mobile-phase flow rate of 0.35 ml/min, column temperature of 45˚C, and injection volume of 2 µl.

MS conditions were mainly the following: The ESI source temperature was 500˚C, with an ion spray voltage of 5,500 V for the positive mode and -4,500 V for the negative mode. The ion source GS I and GS II pressures were 45 psi and 55 psi, respectively, and the CUR pressure was 25 psi. CAD parameters were set to medium values. In the Qtrap, each ion pair was scanned for detection based on the optimized DP and CE.

After MS data were analysed with software Analyst 1.6.3 (AB SCIEX), substances were accurately identified by mass-to-charge ratio (m/z) and RT of multiple ion pairs and second spectra of the identified metabolites using the self-built target standard database MWDB (including secondary spectra and RT), the public MHK database compiled by Metware [comprising Metlin, Human Metabolome Database (HMDB), and KEGG databases and including secondary spectra and RT], and the MetDNA algorithm.

After screening the characteristic ions of each substance by triple quadrupole, the signal strength of characteristic ions in the detector was obtained. Subsequently, through integrating and correcting chromatographic peaks using the software MultiQuant (version 3.0.3; AB SCIEX), the relative content of the corresponding substance represented by the area of each chromatographic peak, was obtained.

Significantly regulated metabolites between groups were determined by variable importance in projection (VIP) ≥1 and absolute Log2FC ≥1.5 or ≤0.67. VIP values were extracted from Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) results, which also contain score plots and permutation plots, and permutation plots were generated using R package MetaboAnalystR (version 1.0.1; https://www.metaboanalyst.ca/). The data was log transformed (log2) and mean centering was performed before OPLS-DA. In order to avoid overfitting, a permutation test (200 permutations) was performed. Metabolite annotation was performed based on the KEGG compound database (http://www.kegg.jp/kegg/compound/).

Lysis solution (8 M urea/100 mM Tris-Cl) was added to the sample. The resulting mixture was sonicated in a water bath and incubated with dithiothreitol (DTT) for 1 h at 37˚C. Next, iodoacetamide (IAA) was added and the alkylation reaction was allowed to take place at room temperature in the dark to block the sulfhydryl groups. Protein concentration was determined using the Bradford method. After protein quantification, 50 µg of the protein sample was used for sodium dodecyl-sulphate polyacrylamide gel electrophoresis (5% concentrated gel; 12% separation gel), which was based on observation of Coomassie brilliant, blue-stained protein bands. After sample reduction and alkylation, 100 mM Tris-HCl was added to the sample to dilute the urea concentration to <2 M. Trypsin was added at an enzyme-to-protein mass ratio of 1:50 and the mixture was incubated overnight at 37˚C for enzymatic protein cleavage, which was terminated the next day by adding 10% TFA. The supernatant was desalted with Sep-Pak C18, suction-dried, and stored at -20˚C for later use.

MS data were acquired using an Orbitrap Exploris 480 mass spectrometer in tandem with an EASY-nLC 1200 LC system (Thermo Fisher Scientific, Inc.). Peptide samples were solubilized by a loading buffer, aspirated by an autosampler, and loaded to an analytical column (75 µm x25 cm, C18, 1.9 µm, 100 Å) for separation. Two mobile phases (mobile phase A, 0.1% formic acid; mobile phase B, 0.1% formic acid with 80% acetonitrile) were used in a gradient. The flow rate of each liquid phase was set to 300 nl/min. MS data were acquired in DIA mode, with each scan cycle consisting of one MS1 scan (R=60K, AGC=3e6, Max IT=30 msec, scan range=350-1,250 m/z) and 40 MS2 scans of variable windows (R=30K, AGC=1,000%, Max IT=50 msec). High field asymmetric waveform ion mobility spectrometry (FAIMS) was operated (CV-45) and the collision energy was set to 30.

Raw DIA data were analysed using software DIA-NN (v1.8) (32) according to the following steps: i) A spectral library was predicted from the Swiss-Prot human database (which was reviewed on March 12, 2021) by using the DL algorithm of DIA-NN; ii) protein and peptide ions were identified from the predicted library using the match between run (MBR) algorithm with the raw DIA data; iii) the protein and peptide ion identifications were filtered at a 1% false discovery rate (FDR) to obtain quantitative proteomic information for subsequent analysis.

A sample reproducibility test was performed using PCA analysis and correlation coefficient analysis (Pearson correlation test) based on the relative quantification results of protein. The mean expression level of a given protein across all biological replicates in one group was divided by the counterpart in the other group, and the ratio was defined as the fold change (FC), with FC <0.67 or FC >1.5 as the threshold for differential expression of proteins in the two groups. A T-test on the FC data, with P<0.05 considered indicative of statistical significance, was performed to identify differentially expressed proteins DEPs. Functional enrichment analysis of DEPs was performed to identify the Gene Ontology (GO; http://www.geneontology.org/) categories and KEGG pathways (clusterProfiler 3.10.1; https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html).

Widely targeted metabolomics and lipidomics were performed on the plasma samples of the CR/PR and PD/SD groups of patients with OSCC to elucidate the metabolomic differences between the two groups. A total of 904 metabolites belonging to the lipid family, including sterol lipid, fatty acyl class, glycerol phospholipids, sphingolipid, and glyceride class were identified by wildly targeted lipids. A total of 1,517 metabolites were detected by TM-widely targeted metabolomics, including bile acids, amino acid, phenolic acids, nucleotide, and its metabolites.

As shown in the volcano plot, a total of 16 downregulated metabolites and 47 upregulated metabolites were detected by widely targeted lipidomics (Fig. 2A). The upregulated differentially expressed metabolites were mainly Carnitine C6:1, TG(16:0 22:6), PC(14:0-20:5), LNAPE(20:5/N-18:1), TG(10:0-12:0_14;0), PC(15:0_20:5), TG(14:0,_14:1_18:1), PI(18: 0,_20:5), PE(20:5_18:1), and PE(16:0_20:5), and the downregulated differentially expressed metabolites were mainly LPI(20:3/0:0), PE(18:2,_19:1), PC(15:0_20:3), Carnitine C10:1-OH, PC(O-18:1_18;2), PC(O-18:0-20:3), TXB2, PE(O-17:1_20:4), PI(18:1_20:2), and PE(P-17:0_20:4; (Fig. 2B). KEGG enrichment analysis of the differentially expressed metabolites showed that these were mainly enriched in the primary bile acid biosynthesis pathway and the bile secretion pathway, as well as the cholesterol metabolism pathway (Fig. 2C). This suggests that the poor prognosis of patients in the SD/PD group may be related to these three pathways.

Moreover, widely targeted metabolomics showed that a total of 74 metabolites were downregulated and 60 metabolites were upregulated (Fig. 3A). The top ten up or downregulated metabolites are shown in Fig. 3B. KEGG enrichment analysis of the differentially expressed metabolites showed that these were mainly enriched in the ‘NF-kappaB signaling pathway’, the ‘primary bile acid biosynthesis’, the ‘cholesterol metabolism’ (Fig. 3C), suggesting that the poor prognosis of patients in the SD/PD group may be related to these pathways.

Quantitative proteomic analysis was performed on the plasma samples of both groups to elucidate the proteomic changes in the SD/PD group. MS data were acquired using the DIA mode, which combined the advantages of traditional shotgun proteomics with those of the selected/multiple reaction monitoring (SRM/MRM) technique, a technique considered as the ‘gold standard’ for MS-based absolute quantification. The entire scan range of the mass spectrometer was divided into several windows by m/z, and all parent ions in each window were then fragmented and detected, followed by collecting and using the fragmentation information of all parent ions for protein characterization and quantification. First of all, data were searched against the Swiss-Prot Human database and the results were filtered at 1% FDR, which identified 12,365 peptides and 1,535 proteins. Second, 113 differentially expressed proteins were identified, comprising 50 upregulated proteins and 63 downregulated proteins (Fig. 4A). For better observation of protein change patterns, proteins with significant differential expression were normalized and a clustered heat map was generated. This showed that protein expression profiles were significantly different between the CR/PR and SD/PD groups (Fig. 4B). Next, all differentially expressed proteins were subjected to enrichment analysis for GO categories using ClusterProfiler (33) and the results are shown in Fig. 4C. KEGG pathway enrichment analysis revealed that differentially expressed proteins were mainly enriched in the ‘PI3K-Akt signaling pathway’, the ‘NF-kappaB signaling pathway’, and the ‘phospholipase D signaling pathway’ (Fig. 4D).

Correlation analysis between proteomic and metabolomic data was performed to elucidate the mutual regulatory relationship between differentially expressed proteins and metabolites. First of all, the common KEGG pathways wherein both differentially expressed proteins and metabolites were enriched were identified according to KEGG pathway enrichment analysis. The combined analysis of widely targeted proteomic and lipidomic data showed that DEMs and DEPs were enriched in the ‘cholesterol metabolism’ pathway (Fig. 5A). The combined analysis of widely targeted metabolomic and proteomic data showed that DEPs and DEMs were enriched in the ‘NF-kappaB signaling pathway’ and in the ‘phospholipase D signaling pathway’ (Fig. 5B). The results indicated that the PD-1 monoclonal antibody resistance may be attributed to the three metabolic signalling pathways.

Considering that cholesterol metabolism, NF-kappaB, and phospholipase D signalling pathways play important roles in PD-1 monoclonal antibody resistance in patients with OSCC, the enrichment of DEMs and DEPs was then analysed in these pathways in detail. The heatmap showed that the expression of triglyceride, cholesteryl ester, and apolipoprotein H (APOH) was upregulated in the SD/PD group (Fig. 6A), indicating that the cholesterol metabolism signalling pathway was correlated with the poor prognosis of patients with OSCC treated with immunotherapy combined with chemotherapy. The expression of IGHV6-1, IGHV3-72, IGHV1-18, IGHV3-11, IGHV2-70, IGHV1-2, LY96, IGHV2-5, KIT and d-myo-inositol-1,4,5-triphosphate enrichment in the NF-kappaB and phospholipase D signalling pathway was also upregulated (Fig. 6B and C), suggesting that the NF-kappaB and phospholipase D signalling pathway was activated to promote the immunotherapy and chemotherapy resistance of OSCC. The process described above indicates that the proteins target metabolites to regulate a series of pathways. Clarifying the correlation between the proteins and metabolites is essential for further screening of the mechanisms of resistance to the PD-1 monoclonal antibody in patients with OSCC. The results showed that TG, CE, glycochenodeoxycholic acid, glycocholic acid, and taurocholic acid were co-regulated by APOH (Fig. 6D and E), indicating that APOH may contribute to the immunotherapy and chemotherapy resistance by regulating the metabolism of TG, CE, glycochenodeoxycholic acid, glycocholic acid, and taurocholic acid. Cholesterol metabolism is described in Fig. 6F.