Extracellular Vesicle-Derived Protein File from Peripheral Blood Predicts Immune-Related Adverse Events in Gastric Cancer Patients Receiving Immunotherapy

Simple Summary Most gastric cancer (GC) patients have already benefited from immune checkpoint inhibitors, but some of them may terminate immunotherapy due to immune-related adverse events (irAEs). Extracellular vesicles have been shown to carry proteins, nucleic acids and other biomacromolecules to recipient cells, which is very important for exploring the potential of biomarkers of irAEs via EV-derived proteins. In 62 GC patients, EV-ICOS and EV-IDO1 were screened from 42 vital proteins as biomarkers of irAEs, and then confirmed in a validating cohort of 40 GC patients. In summary, EV-ICOS and EV-IDO1 can perfectly predict irAEs of ICI treated GC patients. Abstract Immune checkpoint inhibitors (ICIs) initiate a new stage for gastric cancer (GC) therapeutics, and plenty of patients have already benefited from ICIs. Liquid biopsy promotes the development of precision medicine of GC. However, due to the lack of precision biomarkers of immune-related adverse events (irAEs), the safety of ICIs-treated GC patients cannot be guaranteed. In our study, GC patients treated with ICIs were included for investigating the correlation between irAEs of ICIs and corresponding outcomes. We also explored the potential of biomarkers of irAEs via EV-derived proteins. Dynamic plasma was taken from 102 ICIs-treated GC patients generated retrospectively or prospectively, who were divided into discovery and validating cohorts. Plasma EV-derived protein profiles were described, and two EV-proteins, inducible T-cell co-stimulator (EV-ICOS) and indoleamine 2,3-dioxygenase 1(EV-IDO1), from 42 vital proteins were screened to predict the prognosis of ICIs with irAEs. Our work is the first to propose that EV-proteins can predict ICIs-corresponding irAEs, which can be conducive to the diagnosis and treatment of GC patients, and to facilitate the screening of beneficiaries.


Introduction
Gastric cancer (GC) is the fifth most frequently diagnosed cancer and the third leading cause of cancer death worldwide [1]. With the rapid development of immunotherapy [2,3], clinical trials such as checkmate-649 and Attraction-2 have proved the efficacy of immune checkpoint inhibitors (ICIs) in the treatment of gastric cancer [4,5]. As proved by a series of studies, advanced GC patients harboring PD-L1-positivity/high combined positive score (CPS), microsatellite instability (MSI-H)/mismatch repair deficient (dMMR) and Epstein-Barr virus (EBV) infection are more likely to benefit from immunotherapy [6][7][8]. However, the clinical application of ICIs is frequently accompanied with adverse events, such as

Patient Information
Recruitment, treatment, management, follow-up and blood samples of all patients were executed by the Department of Gastrointestinal Oncology, Peking University Cancer Hospital & Institute. A total of 102 GC patients diagnosed with unresectable or advanced GC (including both GEJ (gastroesophageal junction) and non-GEJ types) and administrated with ICI-based regimens from August 2016 to April 2021 were enrolled. All patients had normal function of the heart, liver, lung, kidney and had been ruled out bacterial and viral infections. All patients' irAEs were assessed following Management of Immune Checkpoint Inhibitor-Related Toxicity that was published in 2020 and updated to the NCCN Guidelines [15], and the definition and severity of irAEs were determined according to clinical examinations, biological and imaging data, the response rate was determined by the Response Evaluation Criteria in Solid Tumors (RECIST 1.1 Standards) [16]. Written informed consents was obtained from all donors. All specimens and relevant clinical information were approved by the Institutional Ethics Committee, Peking University Cancer Hospital & Institute for research. This study was conducted in accordance with the Declaration of Helsinki.

Specimen Preparation and Plasma EVs Isolation
Approximately 10 mL peripheral blood was collected with EDTA tube, centrifugated at 4 • C, 3000× g, 10 min for separating plasma and blood cells, then centrifugated at 25 • C, 1800× g, 8 min for obtaining lymphocytes after lysis red blood cells. Extracted blood components were stored at −80 • C for future experiments. EV was isolated from plasma samples by using the SEC-based qEV column (Izon Science, Christchurch, New Zealand), and the centrifugation method (CP100NX, Hitachi, Tokyo, Japan) was also applied in this work for EVs extracting. Finally, EV eluates were filtrated with an Amicon Ultra-4 10 kDa centrifugal filter device (Merck Millipore, Burlington, MA, USA).

EV Expression Array Determination and Identification
In our study, 42 candidate marker proteins, including 19 representative cellular biomarkers in tumor-immune microenvironment, 11 immune checkpoint modulators, 12 therapeutic biomarkers or targets of gastrointestinal cancer, and 4 canonical EV markers, were embraced in the EV expression array. Then the 46 key EV-proteins were printed onto a 3D modified slide surface (Capital Biochip Corp., Beijing, China) using Arrayjet microarrayer (Roslin, UK) to construct a protein chip spectrum. For EV-proteins capture experiment, the protein chip spectrum was initially blocked with 3% BSA (w/v) diluted in PBS for 1 h at room temperature, then incubated with 10 µL unpurified plasma sample that diluted (1:10) in wash buffer (0.05% Tween20 in PBS) sequentially at room temperature on a shake flask at 30 rpm for 2 h, and then incubated overnight at 4 • C. The next day, the slides were washed and incubated with biotinylated detection antibodies (anti-human-CD9, -CD63 and -CD81, LifeSpan BioSciences, Seattle, WA, USA) diluted 1:1500 in wash buffer. After that, Cy3-labelled streptavidin (Life Technologies, Carlsbad, CA, USA) was added in the slides for incubation 1 h at room temperature. After washing, the slides were scanned using the GenePix 4000A microarray scanner (Molecular Devices, San Jose, CA, USA) and GenePix Pro image analysis software (Molecular Devices, CA, USA) was used to analyze the signal intensity of fluorescent images.
Wilcoxon-Mann-Whitney tests were used to analyze expressional diversity among subgroups. Clinical parameters were compared by the Chi-square test. Survival proportions of the irAEs group and non-irAEs group were evaluated by Kaplan-Meier analysis paired with Log-rank test. Only p value < 0.05 was considered statistically significant. All analyses and graphing were performed using SPSS 21.0 software or GraphPad Prism version 6.0.

irAEs May Represent a Better Disease Control Rate of Immunotherapy in GC
As shown in the workflow chart ( Figure 1), 102 GC patients received ICIs-based therapy were enrolled as a 62-patient discovery cohort and a 40-patient validating cohort in our work. In the discovery cohort, 149 time-scaled specimens were collected from 62 GC patients, including 62 specimens from BL (baseline), 32 specimens from CR (complete response)/PR (partial response), 20 specimens from SD (stable disease), and 35 specimens from PD (progressive disease IrAEs' relevance with major clinical parameters were compared by chi-square test. Despite insignificant statistical relevance, all patients received double-ICI therapy developed irAEs. This was in accordance with previous reports, potentially due to double-ICI's hypoactivation of anti-cancer immune responses remarkably igniting irAEs ( Table 2). Compared to the patients without irAEs, patients diagnosed with grade ≥ 1 irAEs displayed comparable ORR (objective response rate) and better DCR (disease control rate) ( Figure S1A). Furthermore, patients who developed grade ≥ 2 irAEs displayed slightly longer, yet statistically insignificant irPFS/irOS (immunotherapy-related progression-free survival/overall survival) than patients without irAEs or with grade ≥ 1 irAEs ( Figure S1B) [17]. These

The Expressional Spectrum of Plasma EV-Proteins Identified ICOS and IDO1 as irAE Predictors
As described in the Materials and Methods section, pre-treatment plasma samples from the discovery cohort were assessed through an antibody-sandwiched protein array to generate the expression profiles of 42 key EV proteins and 4 canonical EV markers ( Figure 2a). We then screened EV proteins associated with irAEs. Among the 42 EV proteins, ICOS and IDO1 displayed higher expression in GC without irAEs than in GC with irAEs ( Figure 2b). The onset, stop, and duration of irAEs of six selected cases are exhibited ( Figure S2). Cases 1-3 have the highest expression of EV-derived ICOS/IDO1, while cases 4-6 have the lowest expression of EV-derived ICOS/IDO1. Patient 1 developed grade 1 diarrhea at week 11 and was alleviated after receiving loperamide. Patient 2 developed grade 2 hypothyroidism at week 11 and took levothyroxine as replacement therapy permanently. Patient 3 developed grade 2 interstitial pneumonitis at week 15 and was cured after a one-month course of methylprednisolone. Patient 4-6 had an onset of grade 1 rashes at week 2, week 3, and week 5, respectively. Patient 5 and patient 6 also suffered from fever and anorexia, and their symptoms gradually improved after receiving nonsteroidal anti-inflammatory drugs. Moreover, ICOS and IDO1 displayed different expression profiles regarding the organs/systems related with irAEs ( Figure 2c). These findings illustrated that EV-derived ICOS and IDO1 may provide clues for the irAEs rather than to predict ICI prognosis in ICI-treated GC patients.

EV-ICOS and EV-IDO1 Dynamically Predicted irAEs
In order to further explore whether EV-ICOS and EV-IDO1 can dynamically predic the occurrence of irAEs, time-scaled blood samples of GC patients were tracked and col lected after baseline therapy. The content of CA199 and CA72.4, two prevalently applied circulating tumor markers in GC, displayed a more rapid reduction in grade irAEs ≥ 1 patients than irAEs = 0 patients along with the treatment (Figure 3a). Similarly, the expres sion values of EV-ICOS and EV-IDO1 were lower in patients with irAEs than in patient without irAEs across dynamic specimens from baseline to the six-month time point afte treatment (Figure 3b). Notably, the time gap from treatment initiation to the occurrence of irAEs was slightly longer in patients expressing high baseline ICOS than in patient expressing low baseline ICOS (3.88 vs. 3.44 weeks), as well as in patients expressing high baseline IDO1 than in patients expressing low baseline IDO1 (4 vs. 3.4 weeks) (Figure 3c) indicating that lower baseline expression of EV-derived ICOS/IDO1 is related to earlie development of irAEs.

EV-ICOS and EV-IDO1 Dynamically Predicted irAEs
In order to further explore whether EV-ICOS and EV-IDO1 can dynamically predict the occurrence of irAEs, time-scaled blood samples of GC patients were tracked and collected after baseline therapy. The content of CA199 and CA72.4, two prevalently applied circulating tumor markers in GC, displayed a more rapid reduction in grade irAEs ≥ 1 patients than irAEs = 0 patients along with the treatment (Figure 3a). Similarly, the expression values of EV-ICOS and EV-IDO1 were lower in patients with irAEs than in patients without irAEs across dynamic specimens from baseline to the six-month time point after treatment (Figure 3b). Notably, the time gap from treatment initiation to the occurrence of irAEs was slightly longer in patients expressing high baseline ICOS than in patients expressing low baseline ICOS (3.88 vs. 3.44 weeks), as well as in patients expressing high baseline IDO1 than in patients expressing low baseline IDO1 (4 vs. 3.4 weeks) (Figure 3c

EV-Derived ICOS/IDO1 Were Irrelevant to Immunotherapeutic Effectiveness, Yet Refected Short-Term Changes of Circulating Tumor Marker CA72.4
We then analyzed the correlation between EV-derived ICOS/IDO1 and immunotherapeutic efficacy. The expression levels of EV-derived ICOS/IDO1 were comparable between patients with or without responses (Figure 4a). When stratifying patients with both irAEs and responses, we found that patients who reached an objective response while overcoming irAEs displayed higher EV-derived ICOS/IDO1 expression levels than other stratifications (Figure 4b), indicating that patients with higher EV-derived ICOS/IDO1 have a better chance to achieve both therapeutic effectiveness and safety. Both EV-derived ICOS and ICO1 were largely irrelevant to irPFS or irOS (Figure 4c). On the other hand, expressions of both EV-derived ICOS and IDO1 displayed a high correlation with the content of circulating tumor marker CA72.4 (Figure 4d). These data suggested that EV-derived ICOS and IDO1 were more likely to be associated with short-term treatment outcome. We then analyzed the correlation between EV-derived ICOS/IDO1 and immunotherapeutic efficacy. The expression levels of EV-derived ICOS/IDO1 were comparable between patients with or without responses (Figure 4a). When stratifying patients with both irAEs and responses, we found that patients who reached an objective response while overcoming irAEs displayed higher EV-derived ICOS/IDO1 expression levels than other stratifications (Figure 4b), indicating that patients with higher EV-derived ICOS/IDO1 have a better chance to achieve both therapeutic effectiveness and safety. Both EV-derived ICOS and ICO1 were largely irrelevant to irPFS or irOS (Figure 4c). On the other hand, expressions of both EV-derived ICOS and IDO1 displayed a high correlation with the content of circulating tumor marker CA72.4 (Figure 4d). These data suggested that EV-derived ICOS and IDO1 were more likely to be associated with short-term treatment outcome.

EV-Derived ICOS/IDO1 Remained as Predictive Biomarkers for irAEs in Prospective Validation
To confirm the conclusion that we obtained in the discovery cohort, we then prospectively enrolled 40 ICI-treated patients and collected their pre-treatment plasma specimens as a validating cohort. Since all patients were alive during follow-up, only irPFS was recorded for analysis. For these 40 patients, 30% (12/40) received ICI monotherapy, 62.5% (25/40) received ICI plus chemotherapy and 7.5% (3/40) received double-ICI therapy, while irAEs were observed in 32.5% (13/40) patients. These irAEs were composed of 20% (8/40) grade 1, 10% (4/10) grade 2 and 2.5% (1/40) grade 3, while for all irAEs cases, 46.1 (6/13) were related with skin, 15.4% (2/13) with digestive system, 15.4% (2/13) with hematopoietic system, 15.4% (2/13) with endocrine system, and 7.7% (1/13) with bone and muscle (Table 3). When comparing irAE's relevance with major clinical parameters, patients who received double-ICI therapy displayed a significantly higher trend to developed irAEs (Table 4), which was in accordance with the discovery cohort.  In the validating cohort, patients with irAEs achieved higher DCR than patients without irAEs ( Figure S3A), while patients who developed grade ≥ 2 irAEs also displayed slightly longer irPFS than patients without irAEs or with grade ≥ 1 irAEs ( Figure S3B). EV-derived ICOS and IDO1 displayed higher baseline expression in patients without irAEs than in patients with irAEs (Figure 5a), which was in accordance with our findings in the discovery cohort. However, the EV-derived ICOS/IDO1 expressional distributions among different organs or systems were distinct in validating cohort and discovery cohort (Figure 5b), suggesting that EV-derived ICOS/IDO1 were insufficient to predict the organs/systems-specific irAEs. Similar to discovery cohort, the time gap from treatment baseline to irAE occurrence was longer in patients expressing high baseline ICOS than in patients expressing low baseline ICOS (7 vs. 2.25 weeks), as well as in patients expressing high baseline IDO1 than in patients expressing low baseline IDO1 (8.25 vs. 2.22 weeks) (Figure 5c). Taken together, these results demonstrated that GC patients bearing higher EV-ICOS or EV-IDO1 had both lower risk and shorter interval to develop irAEs, representing a group of patients with better tolerance to ICIs. baseline to irAE occurrence was longer in patients expressing high baseline ICOS than in patients expressing low baseline ICOS (7 vs. 2.25 weeks), as well as in patients expressing high baseline IDO1 than in patients expressing low baseline IDO1 (8.25 vs. 2.22 weeks) (Figure 5c). Taken together, these results demonstrated that GC patients bearing higher EV-ICOS or EV-IDO1 had both lower risk and shorter interval to develop irAEs, representing a group of patients with better tolerance to ICIs. Similar to discovery cohort, there is no evident correlation between the two EV-proteins and ORR/DCR (Figure 6a), while irAEs patients who reached objective response or disease control displayed higher EV-derived ICOS/IDO1 expression than other stratifications (Figure 6b), indicating that patients bearing high EV-ICOS or EV-IDO1 may achieve better therapeutic responses and less toxicities from ICIs. Additionally, levels of EV-derived ICOS/IDO1 were also positively corelated with the value of plasma CA72.4 ( Figure  6c). Collectively, our work demonstrated that although EV-derived ICOS and IDO1 were largely irrelevant to immunotherapeutic responses or prognosis, they were efficient Similar to discovery cohort, there is no evident correlation between the two EV-proteins and ORR/DCR (Figure 6a), while irAEs patients who reached objective response or disease control displayed higher EV-derived ICOS/IDO1 expression than other stratifications (Figure 6b), indicating that patients bearing high EV-ICOS or EV-IDO1 may achieve better therapeutic responses and less toxicities from ICIs. Additionally, levels of EV-derived ICOS/IDO1 were also positively corelated with the value of plasma CA72.4 (Figure 6c). Collectively, our work demonstrated that although EV-derived ICOS and IDO1 were largely irrelevant to immunotherapeutic responses or prognosis, they were efficient predictive biomarkers of irAEs. Dynamic alternations of these two EV-derived proteins in plasma provide guidance on clinical management of irAEs.

ICOS and IDO1 Were Related with Unique Tumor-Microenvironmental Features
Since EV-derived ICOS and IDO1 exhibited similar predictive roles for GC irAEs, we assessed their expressional profiles and found that these two EV-derived proteins were tightly linked in both discovery and validating cohorts ( Figure 7A). The interaction of two molecules was also observed on tissue-derived transcript levels extracted from gastric cancer datasets TCGA and GSE62254 ( Figure 7B), suggesting that ICOS and IDO1 may exert analogous functions in modulating tumor immune microenvironment (TIME). We next analyzed the TIME cell compositions related with ICOS and IDO1 in TCGA and GSE62254. In both datasets, higher expressions of ICOS/IDO1 on tissue level consistently displayed higher infiltrating immune cells, including B cells, T cells, NK cells, dendritic cells, macrophages, neutrophils and their related subtypes ( Figure 7C,D). These phenomena reflected that ICOS and IDO1 expression in both plasma and tissue levels was reflective of unique tumor-microenvironmental features. However, due to the complexity of

ICOS and IDO1 Were Related with Unique Tumor-Microenvironmental Features
Since EV-derived ICOS and IDO1 exhibited similar predictive roles for GC irAEs, we assessed their expressional profiles and found that these two EV-derived proteins were tightly linked in both discovery and validating cohorts (Figure 7a). The interaction of two molecules was also observed on tissue-derived transcript levels extracted from gastric cancer datasets TCGA and GSE62254 (Figure 7b), suggesting that ICOS and IDO1 may exert analogous functions in modulating tumor immune microenvironment (TIME). We next analyzed the TIME cell compositions related with ICOS and IDO1 in TCGA and GSE62254. In both datasets, higher expressions of ICOS/IDO1 on tissue level consistently displayed higher infiltrating immune cells, including B cells, T cells, NK cells, dendritic cells, macrophages, neutrophils and their related subtypes (Figure 7c,d). These phenomena reflected that ICOS and IDO1 expression in both plasma and tissue levels was reflective of unique tumor-microenvironmental features. However, due to the complexity of the TIME, ICOS/IDO's mechanistic linkage with irAEs should be explored by further validations. the TIME, ICOS/IDO's mechanistic linkage with irAEs should be explored by further validations.

Discussion
Contemporarily, immunotherapy has become a central pillar of cancer treatment in various tumor types, including gastric cancer. Although a great many GC patients benefit from immunotherapy, the accompanying irAEs also disturb clinicians for they impede patients from deriving durable responses. IrAEs are a spectrum of diseases featured with unrestrained autoimmunity and inflammations affecting multiple organs. Though generally manageable, certain varieties of irAEs could be life-threatening. Therefore, it remains challenging to seek predictive biomarkers identifying patients at stake in irAEs to facilitate timely clinical intervention.

Discussion
Contemporarily, immunotherapy has become a central pillar of cancer treatment in various tumor types, including gastric cancer. Although a great many GC patients benefit from immunotherapy, the accompanying irAEs also disturb clinicians for they impede patients from deriving durable responses. IrAEs are a spectrum of diseases featured with unrestrained autoimmunity and inflammations affecting multiple organs. Though generally manageable, certain varieties of irAEs could be life-threatening. Therefore, it remains challenging to seek predictive biomarkers identifying patients at stake in irAEs to facilitate timely clinical intervention.
A series of studies have reported candidate biomarkers correlated with IrAEs. Gene expression profiles, C-reactive proteins, human leukocyte antigens (HLAs) and gut microbiome were found to be capable of predicting irAEs [10,[18][19][20][21][22][23]. On serological level, higher or lower expression of specific genes, cytokines/chemokines and autoimmune antibodies were correlated with the occurrence of irAEs [24][25][26]. It has also been reported that early B cell changes, activated CD4 memory T cell abundance and TCR diversity were associated with irAE development [27,28]. It has been suggested MiRNAs, lncRNAs, and proteins carried by Evs are involved in carcinogenesis and the efficacy of ICIs [29], yet the involvement of extracellular vesicles with irAEs has not yet been elucidated. Our study for the first time reported the correlation between EV-derived proteins ICOS/IDO1 and irAEs in patients with GC.
ICOS and IDO1 are proteins related to immunotherapy; relevant target drugs have been produced [30,31], and several studies support that they were associated with allergic diseases [32]. It has been reported that activation of ICOS + Tregs mediates allergic airway diseases, and targeting the ICOS/ICOS-L pathway may disrupt T follicular helper cell responses and ameliorate allergic asthma by disrupting the disease [33,34]. In addition, IDO1, as an immunosuppressive tolerogenic enzyme, has been reported as biomarkers of immune system activation and childhood allergic diseases [35,36]. As a T cell costimulatory receptor, ICOS has a series of immunoagonist antibodies that could multiply the immune response of T cells and improve the ICIs' efficacy [37]. Moreover, IDO1 plays an important role in tumor immunosuppression, and multiple inhibitors have been evaluated In clinical trials [38,39]. However, we did not find proof that EV-derived ICOS or IDO1 exert an influence on immunotherapeutic responses or prognosis. Thus, our current work did not support the assumption of combining ICIs with ICOS-or IDO1-targeted agents.

Conclusions
Immunotherapy innovates cancer therapy while unpredictable irAEs restrict its clinical practice. In our study, a protein chip spectrum that contains 46 key tumor immune microenvironment molecules and classic targets was invented to screen irAEs related markers. Considering that plasma samples are easily obtained, less consumed, and minimally invasive, our detecting procedure was very convenient and accurate to predict irAEs. In conclusion, our results indicate that EV-ICOS and EV-IDO1 are efficient predictors of irAEs, which could assist the decision-making for clinicians and could reduce unnecessary suffering for GC patients treated with ICIs.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers14174167/s1, Figure S1: Relationship between irAEs and ORR/DCR or irPFS of GC patients in discovery cohort. (A) The relationship between irAEs and ORR/DCR. (B) The relationship between irAEs' level and irPFS/irOS. Figure S2: Dynamic presentation of onset/stop and duration of irAEs in ICOS/ IDO1 low and ICOS/IDO1 high cases. Figure

Data Availability Statement:
The data included in this study were described in supplementary information files.