Tumor-originated exosomal lncUEGC1 as a circulating biomarker for early-stage gastric cancer

The screening strategy for identification of EGC-specific exosomal lncRNAs is illustrated in Fig. 1. Plasma samples from patients diagnosed with stage I GC (n = 10) and healthy individuals (n = 5) (Additional file 1: Table S1) and culture media (CM) from different GCCs (AGS, KATO III, NCI-N87, and Hs 746 T; n = 4) and HPSECs (n = 4) were collected. Then, exosome-enriched fractions were prepared using step-wise centrifugation-ultracentrifugation and discontinuous iodixanol gradient methods. Transmission electron microscopy (TEM) and NanoSight particle tracking were used to characterize and quantify the isolated exosomes (Additional file 2: Figure S1A-D). Moreover, western blotting revealed the presence of specific exosome marker proteins (CD9 and CD63) and the absence of the negative control protein tubulin in the exosome-enriched fractions (Additional file 2: Figure S1E-F).

Next, exosomal RNA was purified and analyzed via RNA sequencing on an HiSeq3000 instrument. Based on the RNA sequencing data, 79 exosomal lncRNAs and mRNAs were up-regulated by more than 2-fold in stage I GC patients (n = 10) compared with healthy individuals (n = 5) (Q < 0.00001) (Additional file 2: Figure S1G, Additional file 3: Table S2). Previous studies have demonstrated that circulating exosomal lncRNA ZFAS1, LINC00152 and lncRNA HOTTIP levels were significantly elevated in GC patients compared with those in healthy controls [9‐11]; however, these lncRNAs were not found in our up-regulated exosomal RNA profile, likely because they are not present in EGC. In our previous study, we could not exclude the possibility that these RNAs are secreted from other cells, including immune and inflammatory cells, which are coincidentally present at primary tissue lesions. Thus, exosomal RNA from the CM of four HPSECs and four GCCs was also analyzed via RNA sequencing. Two hundred eighty-five exosomal lncRNAs and mRNAs were up-regulated by more than 2-fold in culture medium from GCCs (n = 4) compared with HPSECs (n = 4) (Q < 0.00001) (Additional file 2: Figure S1H, Additional file 4: Table S3). We combined the results and found that six genes, ENST00000568893.1 (lncUEGC1), ENST00000378432.1 (lncUEGC2), DNM1P18, GSC, LRRC4B, and PKD1P2, were significantly up-regulated in both of the groups (Fig. 1). Hence, we focused on the two potential EGC-specific exosomal lncRNAs with the highest up-regulated fold changes and basic expression levels among the six genes and named them lncUEGC1 and lncUEGC2 (lncRNA up-regulated in the exosomes of gastric cancer).

To further study the robustness of the RNA-seq data, quantitative PCR (qPCR) was performed to analyze exosomal lncUEGC1 and lncUEGC2 expression in EGC specimens from the validation set (Additional file 1: Table S1) and GCCs. As shown in Fig. 2a and b, the relative plasma exosomal lncUEGC1 and lncUEGC2 expression levels were both significantly up-regulated in stage I and II GC patients (n = 51) compared with healthy controls (n = 60) (fold change > 5, P < 0.0001). Moreover, culture medium from 12 different GCCs and four HPSECs obtained via primary culture from four different healthy donors were used to prepare exosome-enriched fractions. The qPCR analysis results revealed that exosomal lncUEGC1 and lncUEGC2 were significantly up-regulated in GCC (n = 12) culture medium compared with those in HPSEC culture medium (n = 4) (lncUEGC1, fold change = 11.27, P = 0.0024; lncUEGC1, fold change = 4.01, P = 0.0054) (Fig. 2c-d). Collectively, these results were consistent with the RNA-seq data and suggested that lncUEGC1 and lncUEGC2 may have potential as EGC-specific exosomal lncRNAs for early non-invasive GC diagnosis.

Although identification of circulating RNA in plasma is an emerging field for non-invasive diagnostic applications, the majority of long RNA molecules in plasma exhibit poor stability. To examine the stability of lncUEGC1 and lncUEGC2 in plasma, twenty stage I GC specimen plasma samples were directly treated with RNase. The relative plasma levels of total circulating lncUEGC1 were not significantly different between plasma treated with or without RNase (P = 0.4154) (Fig. 2e). However, the relative plasma levels of total circulating lncUEGC2 were down-regulated by approximately two-fold in plasma under RNase treatment (P = 0.0305) (Fig. 2f). We also evaluated lncUEGC1 and lncUEGC2 levels in exosomes-depleted plasma supernatants after ultracentrifugation. The expression of lncUEGC1 were abundant in exosomes instead of exosomes-depleted plasma, but the expression of lncUEGC2 were not significantly different between exosomes and exosomes-depleted plasma (Additional file 5: Figure S2), which revealed that a proportion of lncUEGC2 was not encapsulated within exosomes.

Moreover, we analyzed the correlation between the relative lncUEGC1 and lncUEGC2 levels in plasma exosomal RNA and plasma total circulating RNA. The relative lncUEGC1 and lncUEGC2 levels were both closely related in plasma exosomal RNA and plasma total circulating RNA according to the calculated correlation coefficients (Fig. 2g-h). Taken together, the above results suggest that almost all the plasma lncUEGC1 was encapsulated within exosomes and thus protected from RNase degradation rather than circulating freely in the plasma.

To determine the diagnostic power of exosomal lncUEGC1 and lncUEGC2 and the conventional tumor marker CEA for EGC detection, we examined plasma exosomal lncUEGC1 and lncUEGC2 expression levels and serum CEA levels for their capacity to distinguish EGC patients from those with premalignant CAG lesions as well as from healthy controls. As shown in Fig. 3a-c, the relative plasma exosomal lncUEGC1 and lncUEGC2 expression levels and serum CEA levels were all significantly up-regulated in stage I (n = 23) or II (n = 28) GC patients compared with healthy controls (n = 60), but only plasma exosomal lncUEGC1 relative expression levels were significantly up-regulated in stage I (n = 23) GC patients compared with CAG patients (n = 18).

Furthermore, the receiver operating characteristic (ROC) curve (AUC) was generated to evaluate the diagnostic accuracy. Plasma exosomal lncUEGC1 exhibited an AUC value of 0.8760 (P < 0.0001) in distinguishing stage I and II GC patients (n = 51) from healthy controls (n = 60), while plasma exosomal lncUEGC2 exhibited an AUC = 0.7582 (P < 0.0001) and serum CEA exhibited an AUC = 0.6614 (P = 0.0035) (Fig. 3d, Additional file 6: Table S4). The AUC value of exosomal lncUEGC1 was significantly higher than those of the combined serum tumor markers (CEA, CA19-9, CA72-4 and CA12-5) and potential plasma miRNA biomarkers identified in recent studies of EGC diagnosis [12, 13]. More importantly, our diagnostic accuracy study selectively restricted GC patients to those with stage I and II GC, excluding advanced stages, which indicated that lncUEGC1 may be more convincing for EGC diagnosis than markers identified in previous studies investigating diagnostic accuracy in GC patients at all stages [14‐16]. Similar results were achieved in distinguishing stage I GC patients (n = 23) from healthy controls (n = 60); plasma exosomal lncUEGC1 still showed better diagnostic efficiency, with an AUC = 0.8500 (P < 0.0001), compared with plasma exosomal lncUEGC2, with an AUC = 0.7486 (P = 0.0005), and serum CEA, with an AUC = 0.6424 (P = 0.0456) (Fig. 3e, Additional file 6: Table S4). Because it is difficult to distinguish between early-stage gastric cancer and premalignant lesions using the current liquid biopsy index, we further generated the ROC curve of plasma exosomal lncUEGC1 and lncUEGC2 and serum CEA in distinguishing stage I GC patients (n = 23) from CAG patients (n = 18). As shown in Fig. 3f and Additional file 6: Table S4, plasma exosomal lncUEGC1 exhibited a high AUC value of 0.8406 (P = 0.0002) compared with plasma exosomal lncUEGC2, with an AUC = 0.6522 (P = 0.0980), and serum CEA, with an AUC = 0.6123 (P = 0.2219). Collectively, these results provide evidence that plasma exosomal lncUEGC1 may serve as a primary diagnostic GC biomarker. Moreover, this index might effectively distinguish between EGC and premalignant lesions. All methods and meterials used during this study are included in Additional file 7.