Plasma exosomal DOK3 reflects immunological states in lung tumor and predicts prognosis of gefitinib treatment

Abstract

To identify liquid biomarkers that predict clinical outcomes of epidermal growth factor receptor‐tyrosine kinase inhibitor (EGFR‐TKI), we enrolled patients with EGFR gene mutation‐positive non–small‐cell lung cancer who were intended to receive gefitinib treatment. Using plasma samples obtained prior to gefitinib treatment from 12 enrolled patients, we performed comprehensive proteomic analysis of plasma exosomes to explore proteins correlating with tumor reduction rate (TRR), progression‐free survival (PFS), or overall survival (OS). Of the detected 1769 proteins, 119, 130, or 119 proteins demonstrated a strong correlation (|r| > 0.5) with TRR, PFS, or OS, respectively. Interestingly, 34 (29%), 41 (32%), or 27 (23%) of them, respectively, were functionally involved in the regulation of the immune response. CD8α chain was consistently listed as a molecule positively correlated with PFS and OS, suggesting that the long‐lasting effects of gefitinib may be due to the antitumor effects of CD8⁺ T cells, as well as the induction of immunogenic apoptosis of tumor cells by blocking the EGFR signaling pathway. Notably, Doking Protein 3 (DOK3), a molecule involved in B‐cell receptor signaling, and some immunoglobulin and complement molecules exhibited a clear correlation with PFS longevity of gefitinib treatment. Indeed, the strong expression of DOK3 in B cells was confirmed within tertiary lymphoid structures of lung cancer tissues derived from patients with long PFS. These findings suggest that the patients with active B‐cell and T‐cell immunity as a host immunological feature are more likely to benefit from gefitinib therapy. Circulating exosomal DOK3 has the potential as a predictive marker of response to gefitinib indicating this immunological feature.

DOK3+ B cells in tertiary lymphoid structures (TLS) of lung cancer tissues and DOK3 levels in plasma exosomes showed clear correlation with both PFS and OS of gefitinib treatment.

Abbreviations

DOK3

Docking Protein 3

EGFR

epidermal growth factor receptor

LC/MS

liquid chromatography/mass spectrometry

Lys‐C

lysyl‐endopeptidase

NSCLC

non–small‐cell lung cancer

OS

overall survival

PFS

progression‐free survival

PVDF

polyvinylidene difluoride

TCEP

Tris(2‐carboxyethyl)phosphine

TKI

tyrosine kinase inhibitor

TLS

tertiary lymphoid structures

TRR

tumor reduction rate

1. INTRODUCTION

Gefitinib is a first‐generation EGFR‐TKI that shows significant antitumor effects against EGFR mutation‐positive NSCLC; therapeutic resistance is often acquired with the T790M mutation occurring in approximately half of gefitinib‐resistant cases. ¹ , ² , ³ , ⁴ With the approval of osimertinib for treatment of EGFR‐TKI‐resistant inoperable or recurrent NSCLC with T790M mutation, various changes have occurred in the clinical practice of NSCLC, such as the requirement for re‐biopsy to detect T790M after disease progression or approval of liquid testing for EGFR mutations in resistance cases and at initial diagnosis. ⁵ As a result, the importance of liquid biopsy has been recognized and various studies have been carried out to search for resistant mutations in the EGFR gene using plasma/serum and to elucidate the mechanism of EGFR‐TKI resistance. However, we have not been able to identify patients with a promising long‐term prognosis prior to EGFR‐TKI treatment or those who developed early resistance, implying that limitations still exist in precise prediction of EGFR‐TKI responses using acquired gene mutation profiles.

Exosomes are endosome‐derived vesicles with a diameter of 30–100 nm that are secreted in all biological fluids including blood, urine, saliva, cerebrospinal fluid, and in vitro cell cultures. ⁶ These vesicles form part of a communication system between cells and are thought to be useful as biomarkers for various diseases such as cancer. ⁷ , ⁸ In the research field of cancer diagnosis, exosomes are now attractive targets for biomarker discovery due to their molecular properties. ⁹ In principle, a series of molecules expressed in the original solid tumor cells appears to be detectable as exosomal components in the blood. Based on these characteristics, we aimed to identify predictive indicators associating with the effect of EGFR‐TKI treatment from the patients' plasma exosomes.

In this study, we mainly focused on the novel relationship between immunological activity in the tumor microenvironment and the response to EGFR‐TKI treatment, as a considerable number of immune‐related molecules in circulating exosomes were found to quantitatively correlate with the prognosis of gefitinib‐treated NSCLC patients. In particular, we showed the potential significance of DOK3⁺ B cells in TLS. TLS is a tertiary lymphoid structure identified in a broad spectrum of cancer types, and has recently been reported to correlate with longer prognosis and responsiveness to immune checkpoint blockade (ICB). ⁷ , ⁸ , ¹⁰ , ¹¹ In TLS, salient accumulation and colocalization of B cells, CD4⁺ T cells, and CD8⁺ T cells are observed. Importantly, tumors with TLS have been reported to show a higher density of B‐cell and T‐cell infiltration into the tumors. ⁷

DOK3 is an adapter protein downstream of the B‐cell receptor (BCR). It has been reported that DOK3 is activated in germinal center (GC) B cells after antigen recognition and can attenuate BCR‐mediated calcium signaling to limit the size of GC B cells, maintain optimal PD‐L1, and upregulate PD‐L2 expression to generate long‐lived plasma cells (PCs). ¹² Interestingly, a higher rate of DOK3 knockout mice developed lung adenocarcinoma compared with wild type mice. ¹³ Therefore DOK3 is considered to be a tumor suppressor gene.

Here we provide proteome‐wide conformational information of plasma exosomes in patients with NSCLC and demonstrated potential of exosome‐mediated monitoring of tumor microenvironment immunologic activity as a predictive tool for response to gefitinib.

2. METHODS

2.1. Study design and patients

This study is a single arm observational study with a translational research part (Tables S1 and S2). Briefly, patients were enrolled (1) who had primary lung cancer (advanced or recurrent) with an EGFR mutation (Ex.18, Ex.19, Ex.21), (2) who were under consideration for gefitinib treatment, (3) who had at least one lung lesion with a minimum diameter of 10 mm or at least one lymph node lesions with a minimum diameter of 15 mm (except for lesions previously treated with radiation therapy), and (4) who provided a written consent to participate in the study. Patients were excluded if they had HIV, HBV, or HCV infection, a treatment history of EGFR‐TKI and ALK inhibitor, symptomatic brain metastases, overt interstitial pneumonia/pulmonary fibrosis findings, or active, multiple cancers. This prospective study was approved by the Institutional Ethics Review Board (Teikyo University Review Board 14‐072), and the study was registered on the University Hospital Medical Information Network (UMIN000015830). All participants provided written informed consent prior to the study.

2.2. Treatment

Patients received gefitinib 250 mg/day until disease progression or unacceptable toxicity.

2.3. Endpoint

The primary endpoint is the exploration of plasma biomarkers correlated with gefitinib efficacy (TRR, PFS, and OS). Tumor response was assessed by investigators participating in this study.

2.4. Cell culture

Three human cell lines (A549, NCI‐H23, and Jeko‐1) were obtained from the American Type Culture Collection (ATCC; Manassas, Virginia) and cultured in RPMI 1640 medium (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin G, and 0.1 μg/ml streptomycin. PC‐9 and NCI‐H1975 cells were kindly provided by Dr. Ryohei Katayama and cultured in RPMI 1640 medium supplemented with 10% FBS, 100 U/ml penicillin G, and 0.1 μg/ml streptomycin.

2.5. Exosome isolation from culture medium

Culture medium was replaced with RPMI 1640 medium containing 10% of bovine exosome‐free FBS (Thermo Fisher Scientific, Waltham, Massachusetts). After 48 h culture, 10 ml of culture medium was collected and filtered with a 0.22‐μm filter. Followed by concentration to 1 ml using a 100 kDa cut‐off ultrafiltration tube (Merck Millipore, Burlington, Massachusetts), exosomes were isolated by EVSecond L70 columns (GL Sciences International) according to the manufacturer's instruction.

2.6. Mass spectrometric analysis of plasma exosomes

Exosomes were purified from 200 μl of plasma samples using EVSecond L70 columns (GL Sciences International,) according to the manufacturer's instruction. The vacuum‐dried samples were dissolved and reduced in 1× Laemmli's sample buffer with 10 mM TCEP at 100°C for 10 min. Following alkylation with 50 mM iodoacetamide at ambient temperature for 45 min, protein samples were subjected to SDS‐PAGE. Electrophoresis was stopped at the migration distance of 2 mm from the top edge of the separation gel. After CBB staining, protein bands were excised, destained, and cut finely before in‐gel digestion with trypsin/Lys‐C Mix (Promega) at 37°C for 12 h. The resulting peptides were extracted from gel fragments and analyzed with Orbitrap Fusion Lumos mass spectrometer (Thermo Scientific) combined with UltiMate 3000 RSLC nano‐flow high‐performance liquid chromatography (HPLC) (Thermo Scientific). Peptides were enriched with μ‐Precolumn (0.3 mm i.d. × 5 mm, 5 μm, Thermo Scientific) and separated on AURORA column (0.075 mm i.d. × 250 mm, 1.6 μm, Ion Opticks Pty Ltd,) using a two‐step gradient; 2%–40% acetonitrile for 110 min, followed by 40%–95% acetonitrile for 5 min in the presence of 0.1% formic acid. The analytical parameters of Orbitrap Fusion Lumos were set as follows: resolution of full scans = 50,000, scan range (m/z) = 350–1500, maximum injection time of full scans = 50 ms, AGC target of full scans = 4 × 10⁵, dynamic exclusion duration = 30 s, cycle time of data‐dependent tandem mass spectrometry (MS/MS) acquisition = 2 s, activation type = HCD, detector of MS/MS = ion trap, maximum injection time of MS/MS = 35 ms, AGC target of MS/MS = 1 × 10⁴. The MS/MS spectra were searched against the Homo sapiens protein sequence database (20,366 entries) in SwissProt using Proteome Discoverer 2.4 software (Thermo Scientific), in which peptide identification filters were set at a false discovery rate <1%. Label‐free relative quantification analysis for proteins was performed with the default parameters of Minora Feature Detector node, Feature Mapper node, and Precursor Ions Quantifier node in Proteome Discoverer 2.4 software.

2.7. Statistical analysis

Correlation between exosomal protein abundance at the pretreatment point and objective response rate, PFS, or OS of patients with lung cancer by subsequent gefitinib treatment was evaluated using Pearson's correlation coefficient and its p‐value calculated in R 3.6.2 freeware. Statistical power test and sample size calculation were performed using the pwr library in R.

2.8. Gene ontology analysis

UniProt accession numbers of the targeted proteins were uploaded to the Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resource 6.8. Gene‐annotation enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway mapping were performed on this website.

2.9. Target protein selection

Exosomal proteins with quantitative values in more than half of the samples were used for subsequent analysis. Proteins that showed significant correlation with TRR, PFS, or OS were extracted and further screened for immune‐related molecules from Gene Ontology (GO) annotation information. The cut‐off value of the Pearson's correlation coefficient was set at |r| > 0.5.

2.10. Western blotting

NSCLC patient‐derived plasma samples (200 μl) were fractionated on an EVSecond L70 column according to the manufacturer's instructions. For fraction check analysis, each fraction (100 μl) was dried in a vacuum drier and resuspended in 20 μl Laemmli's SDS sample buffer. To acquire whole exosome‐containing fractions, the second to fifth fractions were combined and subjected to acetone precipitation before resuspension in Laemmli's SDS sample buffer. For analysis of whole cell lysates, cells were lysed with Laemmli's SDS sample buffer at a concentration of 1 mg/ml. Next, 10 μg each of protein sample was separated on 8%–12.5% SDS polyacrylamide gels and transferred onto PVDF membranes (Merck Millipore, #IPVH00010). Following blocking with 4% Block Ace (Yukijirushi Nyugyo, Tokyo, Japan), membranes were incubated with anti‐DOK3 polyclonal antibody (HPA077312, Sigma‐Aldrich), anti‐CD9 monoclonal antibody (SHI‐EXO‐M01, Cosmo Bio), or anti‐calnexin polyclonal antibody (ab22595, abcam). Membranes were then incubated with HRP‐conjugated anti‐mouse IgG (GE Healthcare, Chicago, IL, #NA931‐1Ml) or anti‐rabbit IgG (GE Healthcare, #NA934‐1Ml) and detected using Western Lightning ECL Pro (Perkin Elmer, MA, #NEL121001EA). Quantification of band intensity was performed using Image Lab software version 5.2 (Bio‐Rad Laboratories,).

2.11. Immunohistochemistry of DOK3

Paraffin‐embedded specimens were prepared from patients who showed a high or low level of plasma exosomal DOK3 (Exo‐DOK3). The slides were treated with xylene and ethanol to remove the paraffin. The ENVISION+ Kit/HRP (DakoCytomation,) was used to detect DOK3, CD19, and CD8. After blocking the endogenous peroxidase activity and nonspecific protein staining, affinity‐purified rabbit anti‐DOK3 polyclonal antibody (HPA077312; Sigma‐Aldrich), mouse anti‐CD19 monoclonal antibody (BT51E; Leica Biosystems), or mouse anti‐CD8 monoclonal antibody (NCL‐L‐CD8‐4B11; Leica Biosystems) were added as the primary antibody. Then, slides were treated with horseradish peroxidase (HRP)‐labeled anti‐rabbit IgG for DOK3 staining or anti‐mouse IgG for CD19 and CD8 staining, respectively. Finally, chromogen 3,3′‐diaminobenzidine (DAB) was added as the substrate. Serial sections of the tissue specimens were counterstained with H&E. Quantification of cells with positive staining was performed using ImageJ software. The “Brightness” of “Threshold Color” was set from 111 to 170 to defined DAB‐stained cells for all slides. Then the area of the defined DAB‐stained cells was measured within ~1.7 mm² of tissue sections. Finally, the positive area per unit tissue section (μm²/mm²) was calculated and shown.

3. RESULTS

3.1. Patient characteristics

The 19 patients were enrolled in the study during the 3 years between August 2014 and July 2017. In total, 12 patients (63.16%) were female and 13 patients (68.42%) were nonsmokers; 11 patients (57.89%) had advanced lung cancers and eight patients (42.11%) with postoperative recurrence were registered. Of these eight patients, five received postoperative adjuvant chemotherapy. Seven patients (36.84%) had a history of previous treatment as advanced or recurrent lung cancer, of which six patients had received brain or bone irradiation and one patient had received chemotherapy (CDDP + docetaxel) (Tables S1 and S2).

3.2. Exosomal biomarkers correlating with the efficacy of gefitinib

We first isolated exosomes in plasma samples collected from patients prior to initiation of gefitinib treatment (n = 12) as the discovery cohort (Tables S1 and S2) and subjected them to mass spectrometric proteome analysis (Figure 1). The comprehensive protein identification and label‐free quantification analyses resulted in the detection of 1786 exosomal proteins (Table S3) in which typical exosome marker proteins were highly enriched, such as CD9, CD63, CD81, CD151, ALIX, TSN9, TSN14, and TSN33 (Figure 2). Since these proteins were not detectable directly from crude plasma samples (data not shown), these data indicated that appropriate purification of exosomes was achieved, allowing reliable quantification of exosomal proteins with over 10⁵ of dynamic range.

FIGURE 1.

CONSORT diagram of exosomal biomarker discovery in the AQUA‐GR study. A schematic workflow and experimental parameters for this study are shown as the CONSORT diagram. Briefly, exosomes were isolated from plasma from 12 patients with NSCLC and subjected to comprehensive proteome analysis to explore predictive biomarkers for responses to gefitinib treatment

FIGURE 2.

Quantitative overview of plasma exosomal proteins detected in this study. Protein abundance was calculated based on label‐free relative quantification analysis of LC/MS datasets by Proteome Discoverer 2.4. The distribution for 1769 detected proteins is shown in order of their abundances. The abundance levels of eight typical exosome marker proteins are indicated

Among the 1390 proteins quantified in more than half of the cases, we explored molecules correlated with the efficacy of gefitinib. As the result of correlation analysis between exosomal protein abundances, the TRR, PFS, or OS of patients treated with gefitinib, showed a significant correlation with 119, 130, or 119 proteins, respectively (Pearson's correlation coefficient, |r| > 0.5). Importantly, the subsequent functional evaluation with the GO annotations revealed significant enrichment of immune function‐related proteins within these prognosis‐related factors (Figure 3). Indeed, 34 (29%, Table 1), 41 (32%, Table 2), or 27 (23%, Table 3) of immune function‐related proteins exhibited a statistical correlation with TRR, PFS, or OS, respectively.

FIGURE 3.

Functional assessment of exosomal proteins correlated with clinical outcomes of gefitinib treatment. Molecular functions of 119, 130, or 119 exosomal proteins were assessed by Gene Ontology analysis; these were significantly correlated with TRR, PFS, or OS, respectively, of gefitinib treatment

TABLE 1.

Pearson correlation coefficients of immune‐related protein levels versus TRR for gefitinib treatment

Accession	Description	Pearson's correlation coefficient (baseline level versus ORR)
P30485	HLA class I histocompatibility antigen, B‐47 alpha ch	0.9163
P00736	Complement C1r subcomponent	0.9009
P30460	HLA class I histocompatibility antigen, B‐8 alpha cha	0.8676
Q96FW1	Ubiquitin thioesterase OTUB1	0.8670
P15291	Beta‐1,4‐galactosyltransferase 1	0.8455
P01700	Immunoglobulin lambda variable 1‐47	0.8136
P08571	Monocyte differentiation antigen CD14	0.8001
Q9Y4G8	Rap guanine nucleotide exchange factor 2	0.6760
Q04771	Activin receptor type‐1	0.6438
P10909	Clusterin	0.6290
P04070	Vitamin K‐dependent protein C	0.6130
P18428	Lipopolysaccharide‐binding protein	0.5664
Q08554	Desmocollin‐1	0.5636
Q9UHD2	Serine/threonine‐protein kinase TBK1	0.5578
P09871	Complement C1s subcomponent	0.5475
O14791	Apolipoprotein L1	0.5425
P04114	Apolipoprotein B‐100	0.5397
P47712	Cytosolic phospholipase A2	0.5124
A0A0A0MS15	Immunoglobulin heavy variable 3‐49	−0.6494
Q8IUI8	Cytokine receptor‐like factor 3	−0.5495
P06127	T‐cell surface glycoprotein CD5	−0.5411
Q5JQS6	Germinal center‐associated signaling and motility‐like	−0.5290
P08637	Low affinity immunoglobulin gamma Fc region receptor III‐A	−0.5254
Q8N423	Leukocyte immunoglobulin‐like receptor subfamily B	−0.5199
Q9UIQ6	Leucyl‐cystinyl aminopeptidase	−0.5152
P06331	Immunoglobulin heavy variable 4‐34	−0.5141
P02743	Serum amyloid P‐component	−0.5110

TABLE 2.

Pearson correlation coefficients of immune‐related protein levels versus PFS for gefitinib treatment

Accession	Description	Pearson's correlation coefficient (baseline level versus PFS)
Q7L591	Docking protein 3	0.8783
P01834	Immunoglobulin kappa constant	0.7973
Q9UKG1	DCC‐interacting protein 13‐alpha	0.7722
P06310	Immunoglobulin kappa variable 2‐30	0.7200
P01599	Immunoglobulin kappa variable 1‐17	0.6892
A0A0A0MS15	Immunoglobulin heavy variable 3‐49	0.6626
P00751	Complement factor B	0.6599
P01024	Complement C3	0.6079
P07357	Complement component C8 alpha chain	0.5964
P13671	Complement component C6	0.5891
P07360	Complement component C8 gamma chain	0.5877
A0A0B4J1V0	Immunoglobulin heavy variable 3‐15	0.5701
P01859	Immunoglobulin heavy constant gamma 2	0.5698
Q9BS40	Latexin	0.5688
P01031	Complement C5	0.5660
P10643	Complement component C7	0.5652
P0DOX5	Immunoglobulin gamma‐1 heavy chain	0.5587
P01732	T‐cell surface glycoprotein CD8 alpha chain	0.5584
P07358	Complement component C8 beta chain	0.5467
P06312	Immunoglobulin kappa variable 4‐1	0.5410
P18084	Integrin beta‐5	0.5400
P01619	Immunoglobulin kappa variable 3‐20	0.5325
P28070	Proteasome subunit beta type‐4	0.5285
Q13477	Mucosal addressin cell adhesion molecule 1	−0.6531
P30485	HLA class I histocompatibility antigen, B‐47 alpha ch	−0.6447
Q9Y6Z7	Collectin‐10	−0.6153
P30453	HLA class I histocompatibility antigen, A‐34 alpha ch	−0.5893
P04070	Vitamin K‐dependent protein C	−0.5774
P07225	Vitamin K‐dependent protein S	−0.5691
O00478	Butyrophilin subfamily 3 member A3	−0.5561
P51692	Signal transducer and activator of transcription 5B	−0.5548
P16284	Platelet endothelial cell adhesion molecule	−0.5403
P15907	Beta‐galactoside alpha‐2,6‐sialyltransferase 1	−0.5343
P10909	Clusterin	−0.5339
P16278	Beta‐galactosidase	−0.5329
P08571	Monocyte differentiation antigen CD14	−0.5266
Q9BWP8	Collectin‐11	−0.5242
O14791	Apolipoprotein L1	−0.5202
P11766	Alcohol dehydrogenase class‐3	−0.5135

TABLE 3.

Pearson correlation coefficients of immune‐related protein levels versus OS for gefitinib treatment

Accession	Description	Pearson's correlation coefficient (baseline level versus OS)
Q7L591	Docking protein 3	0.8301
Q9NUQ9	Protein FAM49B	0.7272
P43686	26S proteasome regulatory subunit 6B	0.7108
Q00013	55 kDa erythrocyte membrane protein	0.7033
Q05397	Focal adhesion kinase 1	0.6890
P60900	Proteasome subunit alpha type‐6	0.6596
Q13409	Cytoplasmic dynein 1 intermediate chain 2	0.6520
P20618	Proteasome subunit beta type‐1	0.6261
P06310	Immunoglobulin kappa variable 2‐30	0.6116
O43747	AP‐1 complex subunit gamma‐1	0.5955
P01732	T‐cell surface glycoprotein CD8 alpha chain	0.5815
P28062	Proteasome subunit beta type‐8	0.5796
Q96PL5	Erythroid membrane‐associated protein	0.5484
O43242	26S proteasome non‐ATPase regulatory subunit 3	0.5438
Q13093	Platelet‐activating factor acetylhydrolase	0.5142
O14818	Proteasome subunit alpha type‐7	0.5095
P49720	Proteasome subunit beta type‐3	0.5093
Q13200	26S proteasome non‐ATPase regulatory subunit 2	0.5085
P30492	HLA class I histocompatibility antigen, B‐54 alpha ch	−0.7050
P15291	Beta‐1,4‐galactosyltransferase 1	−0.6564
P30485	HLA class I histocompatibility antigen, B‐47 alpha ch	−0.6509
Q13477	Mucosal addressin cell adhesion molecule 1	−0.6398
O00478	Butyrophilin subfamily 3 member A3	−0.6158
P01042	Kininogen‐1	−0.5938
P15907	Beta‐galactoside alpha‐2,6‐sialyltransferase 1	−0.5840
P30511	HLA class I histocompatibility antigen, alpha chain F	−0.5536
P09871	Complement C1s subcomponent	−0.5447
P10909	Clusterin	−0.5400
P16671	Platelet glycoprotein 4	−0.5322
O75022	Leukocyte immunoglobulin‐like receptor subfamily B	−0.5311
P07225	Vitamin K‐dependent protein S	−0.5143
Q92954	Proteoglycan 4	−0.5142
Q9BWP8	Collectin‐11	−0.5114
Q96FW1	Ubiquitin thioesterase OTUB1	−0.5113

The two molecules most positively correlated with tumor shrinkage following gefitinib treatment were HLA class I histocompatibility antigen B‐47 alpha chain (HLA‐B) and complement C1r subcomponent (C1R) (Table 1). HLA‐B is an HLA class I heavy chain involved in antigen presentation. C1R is known to be involved in the early stages of the canonical complement activation pathway. Notably, the expression levels of these proteins in circulating exosomes predicted the gefitinib‐induced tumor shrinkage prior to treatment (Figure 4). However, HLA‐B was negatively correlated with PFS and OS. Also, C1R was not involved in the prolongation of PFS and OS (Tables 2 and 3), whereas the downstream molecules of C1 in the complement activation pathway (C3, C5, C6, and C8) were positively correlated with PFS (Table 2 and Table S4). Interestingly, T‐cell surface glycoprotein CD8 alpha chain (CD8A) was listed as a molecule positively correlated with both PFS and OS (Tables 2 and 3). These facts suggested that the long‐term prognosis of gefitinib treatment might be affected by molecular context different from that affecting direct tumor shrinkage.

FIGURE 4.

Plasma exosomal proteins that are associated with tumor reduction rate of gefitinib. The abundance of plasma exosomal complement C1r subcomponent (A) or HLA class I histocompatibility antigen B‐47 alpha chain (B) is displayed with the orange line with the y‐axis on the right side. Tumor reduction rate (TRR) of each patient is shown using a blue bar with the y‐axis on the left side. ND, not detected

3.3. DOK3, a novel predictive biomarker for gefitinib treatment outcome

From the correlation analysis between plasma exosomal protein levels and the duration of PFS after gefitinib treatment, Docking protein 3 (DOK3) was determined to be the most highly correlated factor (r = 0.88, Table 2). DOK3 is an adapter protein in the BCR signaling pathway that promotes the differentiation of B cells into PCs¹⁰. Moreover, this protein also demonstrated the strongest correlation with OS after gefitinib treatment (r = 0.83, Table 3). Swimmer plots in Figure 5A,B illustrate a clear concordance of plasma Exo‐DOK3 at the pretreatment time point with PFS and OS, respectively, that interestingly did not correlate with TRR.

FIGURE 5.

Plasma Exo‐DOK3 levels at baseline quantitatively correlate with clinical outcomes of gefitinib treatment. Relationship between Plasma Exo‐DOK3 levels (orange bars with the upper x‐axis) and PFS (A) or OS (B) after gefitinib treatment (blue bars with the bottom x‐axis) is shown for the discovery set (n = 12). The numbers on the y‐axis indicate patient ID. (C, D) Results of plasma Exo‐DOK3 measurement using the independent validation set (n = 7) are displayed as in (A, B) identically. To confirm technological reproducibility, three patient samples from the discovery set (indicated by *) were analyzed simultaneously. (E) Here, 200 μl of plasma sample derived from a patient with NSCLC was loaded onto an EVSecond L70 column. The eluted fractions 1–12 were subjected to western blotting analysis that detected DOK3, CD9, and IgG. IB, immune‐blotting

To evaluate technical and biological reproducibility, Exo‐DOK3 was measured using plasma samples collected from the independently‐prepared validation set (n = 7) and also in three cases selected from the discovery set. For technical reproducibility, the Exo‐DOK3 values of patients ID 003, 009, and 015 in this reanalysis (Figure 5C,D) were statistically consistent with the values in the discovery set analysis (Figure 5A,B), indicating that our mass spectrometric quantification method had sufficient quantitative reliability. More importantly, the Exo‐DOK3 levels from seven patients in the validation set demonstrated a clear correlation with both PFS and OS of these patients (r = 0.87 and 0.90, respectively). From the additional power test (n = 12, α = 0.05), the statistical power (1 − β) was 0.99 or 0.96 for the analysis of PFS or OS, respectively, in the discovery set. In the same manner, statistical power was 0.80 or 0.87 for analysis of PFS or OS, respectively, in the validation set (n = 7, α = 0.05). Conversely, the statistically required sample size to support r = 0.88 or 0.83 was n = 6.78 or 8.08 (α = 0.05, β = 0.2) for analysis of PFS or OS, respectively, in the discovery set. Similarly, the required sample size to support r = 0.87 or 0.90 was n = 6.99 or 6.24 for the analysis of PFS or OS, respectively, in the validation set. These power assessment analyses confirmed the statistical correlation between Exo‐DOK3 levels and the prognosis of gefitinib treatment.

As acquisition of resistance mutation (T790M) of EGFR after gefitinib treatment is a key event that may affect long‐term prognosis, we examined the correlations between posttreatment acquisition of T790M and the following factors, exosomal DOK3, PFS, or OS. As shown in Figure S1, the occurrence of this mutation showed less impact on these factors.

To assess the quantitative reliability of LC/MS analyses, we performed western blotting analysis using purified exosomes from the frozen stock of the identical plasma samples used for LC/MS analyses (Figure S2). Quantitative comparison between LC/MS‐based and western blotting‐based values of plasma exosomal DOK3 showed a statistically strong correlation (R ² = 0.743). These results strongly suggested that the abundance of plasma Exo‐DOK3 at a pretreatment time point could have the great potential to be a predictive biomarker for PFS and OS of gefitinib treatment for lung adenocarcinoma patients. Concerning the expression of DOK3 on exosomes, we confirmed the co‐expression of DOK3 and CD9 in the earlier fractions of size exclusion chromatography using an EVSecond L70 column, indicating that DOK3 was encapsulated into exosomes in plasma, rather than existing as a free protein such as IgG (Figure 5E).

To estimate the origin of DOK3⁺ exosomes, immunohistochemical staining was performed using tissue specimens derived from patients with NSCLC and with long PFS (patient ID 005, 607 days) or short PFS (patient ID 010, 73 days). When serial sections were stained with anti‐DOK3 or anti‐CD19, DOK3 expression was predominantly detected in CD19⁺ B cells (Figure 6C,D). In particular, TLS were often seen in patients with longer PFS, in which high expression of DOK3 was found in TLS‐associated B cells (Figure 6A–D). In contrast, patients who did not respond well to gefitinib did not have clear TLS and showed little to no DOK3 expression (Figure 6E–H). To confirm that DOK3⁺ exosomes are derived from B cells rather than lung cancer cells, we isolated exosomes from the culture medium of B‐cell lymphoma cell line, Jeko‐1, and from four lung adenocarcinoma cell lines, A549 (EGFR‐wt), NCI‐H23 (EGFR‐wt), PC‐9 (EGFR‐exon 19 deletion), and NCI‐H1975 (EGFR‐L858R/T790M). Immunoblotting analysis using 10 μg each of whole cell lysates and these exosomes (Figure S3) showed that expression of DOK3 was observed dominantly in Jeko‐1 cells, whereas four lung cancer cells exhibited much lower expression levels of DOK3. These observations were further quantitatively validated by immunohistochemical staining of three antigens (DOK3, CD19, and CD8) using serial sections of six surgical specimens (Figure S4). Among them, more than one TLS site was found in four cases with better PFS and OS (patient IDs 01, 002, 005, and 016) (Figure S4A–D), whereas no TLS was detected in the remaining two patients (patient IDs 008 and 010) (Figure S4E,F). Importantly, when the frequency of cells with positive staining was quantified (μm²/mm²), markedly higher numbers of DOK3⁺, CD19⁺, and CD8⁺ cells were accumulated in tumor regions of tissues derived from patients with better PFS and OS (patient IDs 01, 002, 005, and 016) (Figure S5) compared with patients with poorer PFS and OS (patient IDs 008 and 010). Therefore, TLS‐mediated immunological activity would have an important role in response to gefitinib, which could be monitored via circulating Exo‐DOK3.

FIGURE 6.

Expression of concentrated samples for DOK3 in B cells accumulated at TLS within NSCLC tissues derived from patients with long PFS. Histological and immunohistochemical analyses were performed to confirm the localization of B lymphocytes and the expression of DOK3 in lung cancer tissues derived from the patients with response and nonresponse to gefitinib therapy. H&E staining in responder tissues showed TLS (boxed in A) within the tumor and the infiltration of immune cells therein (A, B). CD19⁺ B cells are observed mainly inside the TLS (C) where DOK3 is also observed predominantly (D). Conversely, in nonresponders, neither TLS nor the infiltration of immune cells was observed, from which DOK3 expression was not observed (E–H)

4. DISCUSSION

To date, the trend for biomarker research for EGFR‐TKI has been mainly focused on acquired resistant mutations of the EGFR gene in tumors. In addition, the results of previous clinical trials showed that EGFR mutant (EGFRm) NSCLC was very poorly responsive to monotherapy with anti‐PD‐1/PD‐L1 antibodies. ¹¹ Although the therapeutic effects of EGFR‐TKIs on tumor microenvironment (TME) have been discussed previously, ¹⁰ , ¹¹ there has been little discussion on whether host immunity, which can be assessed noninvasively from peripheral blood, is involved in the treatment outcome of EGFR‐TKI. Our results suggested that the host immune system is strongly involved in the therapeutic effects of gefitinib, one of the approved EGFR‐TKIs. Of course, this study is a single cohort study based on a limited number of patients, therefore there are indeed some limitations on the correct interpretation of the results.

One of the interesting aspects of the results is that, among the candidate predictive markers of gefitinib with higher correlation coefficients, the screened predictive markers for tumor shrinkage (e.g., HLA‐B and C1R) and those for PFS/OS (e.g., DOK3 and CD8A) were different. In particular, HLA‐B was positively correlated with TRR, but negatively correlated with PFS/OS. This may occur because HLA‐B expression is suppressed after antigen presentation to T cells to prevent hyperactivation of T‐cell immunity. ¹² The results of several previous meta‐analyses in lung cancer research have shown that TRR is not a surrogate marker for OS, ⁶ , ⁹ , ¹³ and the results of this study were consistent with that clinical fact. The immune system may also be involved in tumor shrinkage, however the immune‐related molecules regulating temporary tumor shrinkage and those affecting long‐term sustainable pharmacological effects appear to be different.

Second, several immunoglobulins and complement molecules were identified among the molecules correlating with PFS/OS (Table S4). In general, when an antibody (IgG or IgM/D) binds to a membrane antigen of a bacterium or immunogenic cell that has invaded the body to form an immune complex, C1 binds to this antibody first, and activates C1 and C4, which in turn activate C2–C8 one after another. ¹⁴ As a result, the complex of C9 is consequently embedded on the membrane into the cell wall (membrane) to puncture the bacterium or cells and kill them. Therefore, immunoglobulins and complement factors are very closely related molecules. However, how molecules of the antibody‐producing plasma cells (PC)s and the complement system are involved in attacking EGFRm NSCLC cells in coordination with the EGFR‐TKI will require further investigation.

DOK proteins are enzymatically inactive adapter or scaffold proteins involved in the recruitment of inhibitory molecules. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ These proteins play an important role as negative regulators of immune receptor signaling in immune cells such as B cells and macrophages. Among them, DOK3 is known as a negative regulator of JNK signaling in B cells through its interaction with INPP5D/SHIP1. ¹⁷ The phenotype of DOK3 knockout mice was revealed to increase the development of adenocarcinoma in lung compared with the wild type (40% versus 7%), suggesting that DOK3 expression may contribute to the suppression of carcinogenesis. ¹³ Knockout mice are thought to be more likely to develop lung cancer because of the loss of DOK3 expression in lung epithelial and progenitor cells, followed by the loss of autonomous regulation of ERK and AKT signaling in the cells. ¹³ The DOK3 knockout mice also showed markedly attenuated PD‐L1 expression in GC B cells, suggesting that DOK3 plays an important role in B‐cell differentiation to PCs by regulating PD‐L1 expression. ¹² Recent reports have revealed that PD‐1 signaling plays an important role in the regulation of T follicular‐helper (Tfh) cell differentiation and function. ¹⁴ , ²² , ²³ As PD‐1 signaling plays an essential role in regulating Tfh cell differentiation and its function, it is possible that Tfh cells are coordinately regulated by PD‐L1, which is regulated by DOK3 in GC B cells or PC during B‐cell differentiation. Further analysis on these regulatory mechanisms will be warranted.

In conclusion, the pharmacological effects of gefitinib were not only due to its direct effect of inducing apoptosis in tumor cells harboring EGFR gene mutations, but also due to the immune response to tumor cells by CD8⁺ T‐cell and B‐cell immunity, suggesting the importance of understanding the immune status of the host. Long‐term efficacy and transient tumor reduction with gefitinib treatment may be determined by different immunological factors, and the magnitude of the treatment effect may be predictable prior to administration in the future. The most promising predictor of long‐term efficacy of gefitinib is DOK3, a molecule involved in BCR signaling. The validity of DOK3 as a predictive biomarker for the durable response to EGFR‐TKIs warrants further investigation in an independent cohort.

AUTHOR CONTRIBUTIONS

Conception, design, and study supervision: K.U. Clinical sample collection: K.H., H.Y., T.I., T, H., T.S., N.M., N.S., and M.K. Experiments and data curation: R.O., R.F., N.S., and H.S. Writing of the manuscript: R.O. and K.U.

FUNDING INFORMATION

This study was supported by the research fund from Nippon Boehringer Ingelheim Co., Ltd.

DISCLOSURE

N.S. and K.U. received funding support for this study from Nippon Boehringer Ingelheim Co., Ltd. The other authors have no conflict of interest to be disclosed. K.U. is an associate editor of the journal.

ETHICAL APPROVAL

Approval of the research protocol by an Institutional Reviewer Board: This prospective study was approved by the Institutional Ethics Review Board (Teikyo University Review Board 14–072). Informed Consent: All participants provided written informed consent prior to the study. Registry and the Registration No. of the study/trial: This study was registered on the University Hospital Medical Information Network (UMIN000015830). Animal Studies: N/A.

Supporting information

Figure S1

Figure S2

Figure S3

Figure S4

Figure S5

Table S1‐S4

Ochiai R, Hayashi K, Yamamoto H, et al. Plasma exosomal DOK3 reflects immunological states in lung tumor and predicts prognosis of gefitinib treatment. Cancer Sci. 2022;113:3960‐3971. doi: 10.1111/cas.15512