Plasma extracellular vesicle transcriptomics identifies CD160 for predicting immunochemotherapy efficacy in lung cancer

Abstract Better biomarkers are needed to improve the efficacy of immune checkpoint inhibitors in lung adenocarcinoma (LUAD) treatment. We investigated the plasma extracellular vesicle (EV)‐derived long RNAs (exLRs) in unresectable/advanced LUAD to explore biomarkers for immunochemotherapy. Seventy‐four LUAD patients without targetable mutations receiving first‐line anti‐programmed cell death 1 (PD‐1) immunochemotherapy were enrolled. Their exLRs were profiled through plasma EV transcriptome sequencing. Biomarkers were analyzed against response rate and survival using pre‐ and post‐treatment samples in the retrospective cohort (n = 36) and prospective cohort (n = 38). The results showed that LUAD patients demonstrated a distinct exLR profile from the healthy individuals (n = 56), and T‐cell activation‐related pathways were enriched in responders. Among T‐cell activation exLRs, CD160 exhibited a strong correlation with survival. In the retrospective cohort, the high baseline EV‐derived CD160 level correlated with prolonged progression‐free survival (PFS) (P < 0.001) and overall survival (OS) (P = 0.005), with an area under the curve (AUC) of 0.784 for differentiating responders from non‐responders. In the prospective cohort, the CD160‐high patients also showed prolonged PFS (P = 0.003) and OS (P = 0.014) and a promising AUC of 0.648. The predictive value of CD160 expression was validated by real‐time quantitative PCR. We also identified the dynamics of EV‐derived CD160 for monitoring therapeutic response. The elevated baseline CD160 reflected a higher abundance of circulating NK cells and CD8+‐naïve T cells, suggesting more active host immunity. In addition, increased CD160 levels of tumors also correlated with a favorable prognosis in LUAD patients. Together, plasma EV transcriptome analysis revealed the role of the baseline CD160 level and early post‐treatment CD160 dynamics for predicting the response to anti‐PD‐1 immunochemotherapy in LUAD patients.


| INTRODUC TI ON
Lung cancer has been the leading cause of cancer-related mortality worldwide for many years. Lung adenocarcinoma (LUAD) represents the most common histological subtype, accounting for approximately 40% of lung cancers. 1,2 In recent years, immune checkpoint inhibitors (ICIs) targeting programmed cell death 1 (PD-1) and its ligand (PD-L1) have revolutionized the treatment of advanced nonsmall cell lung cancer (NSCLC) without targetable genetic mutations. 3 The PD-1 inhibitor has demonstrated great clinical efficacy in the first-line treatment of advanced LUAD when administered as monotherapy in patients with PD-L1 tumor proportion score ≥50% or in combination with platinum-based chemotherapy regardless of tumor PD-L1 expression. 4,5 Biomarkers predicting immunotherapy response for lung cancer include the expression of PD-L1, tumor mutational burden (TMB), and microsatellite instability (MSI)/mismatch repair deficiency (dMMR). 4,[6][7][8] However, NSCLC patients with low tumor PD-L1 expression or TMB may still benefit from PD-1 inhibitor plus chemotherapy. 5,9,10 MSI/dMMR is rare in lung cancer, with an incidence of less than 1%. 11 However, difficulties are often experienced in collecting multiple, sufficient lung tissue biopsies for use as tissuebased biomarkers. Due to the intra-tumoral spatial heterogeneity, single biopsy specimens may not be indicative of the whole tumor microenvironment (TME) and the systemic antitumor immune response. 12 Furthermore, the wide application of immunochemotherapy underscores the importance of new biomarker discovery, as studies have revealed that the predictive effectiveness of known ICI-related markers, such as PD-L1 and TMB, in LUAD weakens after the addition of chemotherapy to ICI. 5,13 Therefore, improved predictive biomarkers are needed to better stratify patients for their response to immunochemotherapy. Extracellular vesicles (EVs), mainly including exosomes and microvesicles, are lipid-bilayered nanoparticles secreted from most cell types into the peripheral circulation. 14,15 Cargos carried by EVs contain different nucleic acids, proteins, and lipids, which may represent the parental cells of the EVs and mediate the functions of recipient cells. 16 Recently, EV-based liquid biopsy has gained increasing attention in the early diagnosis, progression monitoring, prognosis prediction, and therapy efficacy assessment of cancer. [17][18][19] Jin et al. reported that plasma EV-derived miRNAs could be used to distinguish LUAD from lung squamous cell carcinoma (LUSC) in the early diagnosis of NSCLC. 20 de Miguel-Perez et al. reported that EV PD-L1 dynamics could predict clinical response to ICIs in NSCLC patients. 21 EV-derived long RNAs (exLRs) mainly comprise messenger RNA (mRNA), circular RNA (circRNA), and long non-coding RNA (lncRNA). 22 Our previous studies have demonstrated exLRs are promising diagnostic and prognostic biomarkers of pancreatic ductal adenocarcinoma and breast cancer. [23][24][25] However, the potential role of exLRs in the ICI-based treatment of advanced NSCLC has yet to be investigated.
In this study, we profiled the plasma exLRs in locally advanced/ metastatic LUAD patients without targetable oncogenic drivers and explored the potential of exLRs as predictive biomarkers to stratify patients for receiving first-line immunochemotherapy of the anti-PD-1 agent. Notably, we found that EV-derived CD160 may predict the immunochemotherapy efficacy in LUAD. higher abundance of circulating NK cells and CD8 + -naïve T cells, suggesting more active host immunity. In addition, increased CD160 levels of tumors also correlated with a favorable prognosis in LUAD patients. Together, plasma EV transcriptome analysis revealed the role of the baseline CD160 level and early post-treatment CD160 dynamics for predicting the response to anti-PD-1 immunochemotherapy in LUAD patients.

K E Y W O R D S
extracellular vesicle, immunochemotherapy, lung adenocarcinoma, RNA sequencing, therapeutic biomarker Blood samples were collected from healthy donors (n = 56), all LUAD patients before the first administration of immunochemotherapy (baseline, n = 74), and the retrospective cohort during treatment (dynamic post-treatment, n = 46). For dynamic post-treatment plasma specimens, 36 out of 46 samples were collected from the 36 patients after two cycles of immunochemotherapy, including five patients who developed PD after the first response evaluation.
The remaining 10 samples were collected after four cycles of immunochemotherapy (n = 4) and at the time of disease progression (n = 6), respectively, except for patients who developed PD at the first response evaluation.

| Response evaluation
Response was evaluated according to the Response Evaluation Criteria in Solid Tumors version 1.1. 27 We defined progressionfree survival (PFS) time as the date from the initiation of immunochemotherapy until either the occurrence of disease progression or death from any cause. Overall survival (OS) was defined as the date from the initiation of immunochemotherapy until the date of death from any cause. Patients who had not progressed or died at the time of the data cutoff date or who were lost to follow-up before progression or death at the time of their last contact were censored. The objective response rate (ORR) was defined as the proportion of patients with a complete response (CR) or partial response (PR) as the best overall response. Patients who achieved CR or PR as the best overall response during immunochemotherapy with PFS ≥6 months were defined as responders. Patients who achieved stable disease (SD) or progressive disease (PD) as the best overall response or patients with PFS <6 months were defined as non-responders.

| Plasma separation, extracellular vesicle purification, and extracellular vesicle-derived long RNA
Briefly, the EV and RNAs were isolated using an exoRNeasy Serum/ Plasma Kit (Qiagen). Plasma separation, EV purification, and characterization (including transmission electron microscopy, size distribution measurement, and western blots), and exLR isolation are described in detail in the Supporting Information and carried out as previously described. 22,23

| Extracellular vesicle-derived long RNA sequencing analysis
Extracellular vesicle-derived long RNAs isolated from plasma were treated with DNase I (NEB) to remove DNA. Strand-specific RNA-seq libraries were then prepared using the SMARTer Stranded Total RNA-seq kit (Clontech). Library quality was analyzed using a Qubit Fluorometer (Thermo Fisher Scientific) and Qsep100 (BiOptic); 2 × 150 bp paired-end sequencing was performed on an Illumina sequencing platform. The raw sequencing reads were filtered by FastQC (version 0.11.8) and aligned to the Gencode human genome (GRCh38) using the read aligner STAR (version 2.7.1a). 28 Gene expression levels were then quantified by featureCounts (version 1.6.3) 29

| Patient overview
This study enrolled a total of 74 LUAD patients and 56 healthy individuals and (Table S1). Among the LUAD patients (Table S2), the median age was 63 years in the retrospective cohort (n = 36) and 65 years in the prospective cohort (n = 38). In both cohorts, most patients were male, previous or current smokers, in stage IV and common metastatic sites in the chest and bone. Eleven (19%) patients in the retrospective cohort had brain metastasis, while three (7.9%) patients in the prospective cohort had brain metastasis. The PD-L1 status is unknown for 36.1% and 50% of the retrospective and prospective cohorts, respectively. There were no significant differences between the two cohorts in terms of age, gender, stage, metastatic sites, or PD-1 inhibitors received (Table S2)

| Extracellular vesicle-derived long RNA profiling in lung adenocarcinoma patients
To illustrate the exLRs profile of LUAD patients, we isolated RNA from baseline plasma EVs of 74 LUAD patients and 56 healthy donors and implemented exLR-seq for each plasma sample ( Figure 1A).
The isolated vesicles were rounded, cup-shaped, and doublemembrane-bound, as observed by transmission electron microscopy ( Figure 1B). Flow cytometry of the isolated vesicles revealed a heterogeneous population of spherical nanoparticles, with a mean diameter of 100.2 ± 37.1 nm ( Figure 1C). Western blot analysis confirmed the enriched expression of EV markers CD63 and TSG101 in isolated vesicles but not in peripheral blood mononuclear cells (PBMCs).
Meanwhile, the endoplasmic reticulum (ER) marker Calnexin, which is expected to be absent in EVs, was detected in PBMCs but not in isolated vesicles ( Figure 1D).
Approximately 15,000 and 16,000 annotated exLR genes, including mRNAs, lncRNAs, pseudogenes, and circRNAs, were constantly detected in each sample of healthy donors and LUAD patients, respectively. Among the detected exLRs, most are mRNAs. The numbers of enriched mRNAs, pseudogenes, and circRNAs in LUAD patients were greater than those in healthy donors ( Figure 1E). We identified 496 and 378 significantly upregulated and downregulated exLR genes, respectively, in LUAD patients compared with healthy donors (FDR < 0.05, FC > 2) ( Figure 1F). KEGG pathway analysis revealed that differentially expressed exLRs were significantly enriched in cancer-associated pathways, such as transcriptional misregulation in cancer and NF-kappa B signaling pathway ( Figure 1G).

| Baseline extracellular vesiclederived CD160 level correlated with response to immunochemotherapy in lung adenocarcinoma patients
To identify exLR biomarks associated with treatment outcome, we first compared the baseline exLR expression profiles of responders and non-responders in the retrospective group. A total of 498 DEGs (P < 0.05, FC > 1.5) were identified ( Figure 2B). Gene ontology biological process (GO-BP) enrichment analysis indicated that the Tcell activation-related terms were significantly enriched in the DEGs  Figure 2E). In addition, in the prospective cohort, elevated CD160 levels also demonstrated a strong correlation with better PFS although not statistically significant (P = 0.06, Figure S1B).
To further identify the optimal cutoff value of CD160 expression to classify patients' clinical benefit, we used the "maxstat" R package to divide the patients into baseline CD160 expression high and low subgroups by using the optimal cutoff value of 9.779. We also checked if the EV-derived CD160 in LUSC (n = 30) patients treated with anti-PD-1 immunochemotherapy could resemble the predictive function in LUAD. The expression of baseline CD160 was not significantly different between responders and nonresponders ( Figure S4A). The ROC curve of CD160 distinguishing responders from non-responders also showed unsatisfactory predictive capability ( Figure S4B). We did not observe the survival difference (PFS and OS) between CD160-high and CD160-low groups (stratified by the median CD160 expression value) ( Figure S4C,D).
These data indicated that the expression level of baseline EVderived CD160 for predicting the immunochemotherapy efficacy might be cancer type-specific.

| Monitoring the dynamic change of extracellular vesicle-derived CD160 for immunochemotherapy outcomes
The pre-treatment (pre-tx) and post-treatment (post-tx) blood samples from the LUAD patients of the retrospective cohort were used for paired exLR analysis. Notably, we found that the expression level of EV-derived CD160 in the responders significantly decreased from the baseline to early during treatment (the time point of the first imaging evaluation, approximately 4-6 weeks after the first administration of immunochemotherapy, P = 0.002, Figure 4A). However, such changes in CD160 expression were not observed in non-responders (P = 0.56, Figure 4A). Then, we divided all the retrospective patients into CD160-decrease and CD160-increase groups for analysis.

| Extracellular vesicle-derived CD160 interaction with other biomarkers in predicting response to immunochemotherapy
We further explored the interaction between EV-derived CD160 and other biomarkers using univariate and multivariate Cox regression analyses (Table 1). High NLR, high PLR, and high LDH are known immune-based biomarkers associated with inferior survival outcomes in NSCLC patients receiving immunotherapy. 31,32 The univariate analysis showed that the presence of bone metastases, high levels of NLR, and low baseline EV-derived CD160 expression were significantly associated with shorter PFS and OS, but the PD-1 inhibitor that patients received (pembrolizumab or camrelizumab) did not correlate with outcomes (Table 1). Aside from the presence of bone metastases and NLR, the multivariate analysis indicated baseline level of CD160 is also an independent prognostic factor for both PFS (P = 0.019) and OS (P = 0.027) ( Table 1). We also assessed the predictive performance of baseline EV-derived CD160 in all LUAD patients with PD-L1 status available (n = 42). No significant difference in baseline EV-derived CD160 expression was seen among different PD-L1 expression groups ( Figure S5A). Despite the restricted cohort size, improved PFS and OS were still observed in the high CD160 groups across all PD-L1 categories ( Figure S5B-G). In addition, both univariate and multivariate analyses showed a strong correlation between CD160 and PFS/OS, while the PD-L1 status failed to exhibit significant association with either outcome (Table S3). Together, those findings indicated the predictive role of EV-derived CD160 in LUAD, which is independent of the PD-L1 status ( Figure S5).
When the CD160 expression level was used alone or in combination with other prognostic factors (high EV-derived CD160 level, low NLR level, and no bone metastasis), the ORRs (proportions of CR or PR as the best response) ( Figure 5A) and duration of response of ≥6 months ( Figure 5B) were all at high levels. Combining the baseline EV-derived CD160 expression level and NLR level from the hematological testing, we leveraged the predictive power of the resultant model (AUC = 0.733) for all enrolled LUAD patients ( Figure 5C).
To establish a predicting model applicable in clinical use, we constructed a nomogram integrating the presence of bone metastasis, NLR level, and EV-derived CD160 level for 6-month PFS prediction by multivariate Cox regression in all enrolled LUAD patients ( Figure 5D). We confirmed the discrimination degree, concordance, and clinical usefulness of the nomogram for its potential application in the patient's risk stratification ( Figure S6). Collectively, our results suggested EV-derived CD160 as a potential prognostic factor that could leverage the predictive performance for clinical outcomes in patients with unresectable and metastatic LUAD.

| Potential mechanism of extracellular vesiclederived CD160 regulation and lung adenocarcinoma patients' prognosis
We next attempted to understand the mechanism underlying the prognostic role of EV-derived CD160. An EV deconvolution approach, named EV-origin, was developed in our previous study to infer the tissue-cellular source contributions of EVs from the exLR-seq profiles. 33 Here, we utilized EV-origin to decipher the cellular origin heterogeneity of the circulating EVs between CD160-high and CD160-low expression cohorts. Compared to the CD160-low subset, the CD160-high subset showed the most significant enrichment of EVs originating from natural killer (NK) cells and CD8 + -naïve T cells (Figures 6A,B, and S7A), which implied a higher abundance of innate and adaptive immunity components in the CD160-high population. Furthermore, compared to non-responders, responders also tend to have a significantly higher EV-derived CD160 RNA levels from CD8 + -naïve T cells and NK cells ( Figure S7B).
We further referred to the tissue-derived CD160 from the Cancer Genome Atlas (TCGA)-LUAD datasets to assess the concordance of CD160 expression between EVs and tissues in LUAD for predicting clinical outcomes. Overall, the CD160 expression was significantly lower in LUAD tissues than in noncancerous tissues ( Figure 6C). Notably, the patients with a baseline high tissue CD160 level (above the third quartile) displayed a better prognosis than those with a baseline low tissue CD160 level (below the Abbreviations: HR, hazard ratio; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio. first quartile) ( Figure 6D), resembling the better response to immunochemotherapy observed in LUAD patients having a higher baseline EV-derived CD160 level. In addition, we analyzed the TCGA-LUSC datasets. Similar to the results of LUSC EVs, we observed no significant difference in overall survival between the low (below the first quartile) and high (above the third quartile) tissue CD160 level groups ( Figure S8).

| DISCUSS ION
To the best of our knowledge, this is the first study to implement a comprehensive analysis of plasma EV transcriptome in unresectable/advanced LUAD patients and identify predictive biomarkers for immunochemotherapy efficacy. From the transcriptome profiling of plasma EVs, we found that elevated baseline EV-derived CD160 is associated with better clinical outcomes in LUAD patients treated with anti-PD-1 immunochemotherapy, independently of other prognostic factors.
CD160, also known as BY55, is a member of the immunoglobulin superfamily mainly expressed on circulating NK cells and T cells. 34,35 There are different forms of CD160, including the glycosylphosphatidylinositol-anchored, transmembrane, and soluble forms. 36,37 The regulatory role of CD160 on NK cells and T cells is complex and can be either stimulatory or inhibitory depending on the cell type and CD160 form. CD160 could bind to classical and non-classical major histocompatibility complex-I molecules with low affinity to trigger NK cell cytotoxicity and cytokine production. 34,38 It may also co-stimulate with anti-CD3 antibodies to promote proliferation for activated CD8 + T cells. 39  For instance, CD160-herpesvirus entry mediator is an immunesuppressive checkpoint. 34 The high-affinity interaction between them has been shown to inhibit CD4 + T-cell activation and negatively regulate TCR-mediated CD8 + T-cell signaling independently of PD-1 expression. 42,43 The CD160 antibody and PD-1 blockade combination could also enhance human immunodeficiency virusspecific T-cell responses. 37 However, another study showed that the higher baseline expression of CD160 on the CD8 + cytotoxic T-cell subset correlates with a better response to antiretroviral therapy in patients with HIV infection. 44 In addition, it has been reported that CD160 + CD8 + T cells displayed significantly higher cytotoxic function and proliferation than CD160 − CD8 + T cells in chronic virus infection. 35 Thus, the above described discrepancies in the field warrant further investigations to comprehensively understand CD160 in the immune response.
The EV-origin analysis of pre-tx exLR revealed significantly more enriched fractions of EVs originating from NK cells and CD8 + -naïve T cells in CD160-high patients. We speculated that CD160 is an antigen-independent marker of general immune activation here, and the high EV-derived CD160 level could indicate an immune-active TME from the perspectives of NK cells and CD8 + -naïve T cells. Thus, the patients with higher pre-tx EV-derived CD160 levels benefited more from the PD-1 inhibition, as the PD-1/PD-L1 axis blockade would unleash the responses of NK cells and T cells in NSCLC. 45,46 For the CD160 decrease during the treatment in responders, we hypothesized that the expression change might reflect the reduction of inhibitory receptors, including CD160, upon the reinvigoration of exhausted T cells by PD-1 blockade. Additionally, a previous study has identified plasma EVs as a source of CD160 that could be taken up by T cells in patients with chronic lymphocytic leukemia. 47 In our study of LUAD, the high level of baseline EV-derived CD160 is likely to facilitate the therapeutic effect of PD-1 blockade, suggesting CD160 to be a potential target for immune intervention in LUAD patients. However, further investigation is needed to determine the origin of the EV-carried CD160.
We acknowledge several limitations of this study. First, this is a proof-of-principle study performed at a single center with a limited number of participants and a relatively short follow-up time in the prospective cohort. Thus, multicenter prospective studies with a larger sample size and a sufficiently long follow-up time would further verify our findings. Second, due to the multifaceted role of CD160, the underlying mechanism remains unclear. Our findings suggested that the high EV-derived CD160 level may reflect a more immune-active microenvironment, particularly attributed to NK cells and naïve T cells. However, it remains undetermined whether a general pro-immune activation mechanism is conferred by CD160 across different cancer modalities. Future research using larger sample sizes and broader cancer type portfolios with multi-omics analyses is warranted to provide more insight into the role of CD160.

F I G U R E 6
Tumor-derived CD160 level is also associated with favorable prognosis in lung adenocarcinoma (LUAD) patients. Violin plots showing that high baseline extracellular vesicle (EV) CD160 expression is significantly correlated with high abundance of (A) circulating natural killer (NK) cells and (B) circulating CD8 +naïve T cells. (C) Boxplot showing that the CD160 expression in the cancerous tissue is significantly lower than in the adjacent normal tissue in the TCGA-LUAD dataset. (D) Overall survival (OS) of the TCGA-LUAD patients without targetable mutations in the CD160 high (top 25%) and CD160 low (bottom 25%) groups.

ACK N OWLED G M ENTS
The authors thank all the patients and healthy volunteers who participated in this study.

CO N FLI C T O F I NTE R E S T S TATE M E NT
Cuicui Liu and Qiuxiang Ou are employees of Nanjing Geneseeq Technology, China. The remaining authors have no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and analyzed and materials used during the current study are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
Approval of the research protocol by an Institutional Reviewer Board: The study was approved by the Institutional Review Board of the Fudan University Shanghai Cancer Center (Approval No. 2004216-  and was performed in compliance with the Declaration of Helsinki. Informed consent: All informed consent was obtained from the subject(s) and/or guardian(s).
Registry and the registration no. of the study/trial: N/A.