Circular RNA profiles in AHH-1 cells exposed to 0, 2, and 5 Gy 60Co γ-rays. To investigate the effect of ionizing radiation on the expression level of circRNAs in AHH-1, the expression profiles of circRNAs in AHH-1 cells, which exposed to different doses, were identified via transcriptome sequencing. If the circRNA reads were found in at least two parallel samples, this circRNA was detected. Based on this, 883 circRNAs were quantitated in the 0 Gy group. And 1116 circRNAs and 1248 circRNAs were quantitated in the 2 Gy and 5 Gy groups. Of them, 505 circRNAs were identified both in 0 Gy and 2 Gy groups (Fig. 1a). The average length of these circRNAs was 806 nt, consisting of 93.8% ecircRNAs, 4.2% eiciRNAs, and 2.0% ciRNAs. Five hundred and forty-five circRNAs were identified both in 0 Gy and 5 Gy groups (Fig. 2a). The average length of these circRNAs was 785 nt, consisting of 94.3% ecircRNAs, 4.4% eiciRNAs, and 1.3% ciRNAs.
Seven DE-circRNAs were identified in the 2 Gy group compared with the 0 Gy group, consisting of 3 up-regulated and 4 down-regulated circRNAs (Fig. 1b, 1c) (fold-change > 1.5, P < 0.05). These DE-circRNAs consisted of 6 ecircRNAs and 1 ciRNA (Fig. 1d). Eight DE-circRNAs were identified in the 5 Gy group compared with the 0 Gy (Fig. 2b, 2c). In these DE-circRNAs, the expressions of 5 circRNAs were up-regulated and 3 circRNAs expressions were down-regulated. All the DE-circRNAs in the 5 Gy groups were ecircRNAs (Fig. 2d). To sum up, 14 DE-circRNAs were screened in radiation groups. The expression levels of hsa_circZFAND6_008 in both radiation groups were lower than that of in the control group (P < 0.05).
Validation of radiation DE-circRNAs. To verify whether the expression levels of the 14 circRNAs candidates selected by transcriptome sequencing were affected by ionizing radiation, real-time PCR was used for validation. Because of the specific molecular structure of circRNAs, the amplified primers need to span the back junction site and back-to-back. Since circRNAs come from protein-encoding genes, their sequence shows strong similarity to the corresponding mRNA of the same host genes. To avoid amplifying linear transcripts of circRNAs, the specificity of the primers was therefore tested before performing the validation experiments. The products amplified by circRNA primers or mRNA primers were quantified in all samples digested with or without RNase R (RNase R + and RNase R-). As a result, the relative expressions of products amplified by circRNA primers (except for hsa_circZDHHC21_004) were not statistically down-regulated between RNase R + groups and RNase R- groups (Fig. 3a), whereas the expression of the corresponding mRNAs decreased significantly with RNase R digestion (P < 0.01) (Fig. 3b). It indicated that the template of products amplified by the designed back-to-back primers were indeed circRNAs. The back-to-back primers can be used for subsequent validation experiments. Although the expression of hsa_circZDHHC21_004 decreased significantly after RNase R treatment (P < 0.05), the decrease fold was significantly smaller than that of the cognate linear transcript. So the primers for hsa_circZDHHC21_004 could also be used for subsequent validation experiments.
At 4 h and 24 h after irradiation, the relative expressions of 14 candidates were detected. At 4h post exposed to 2 Gy, hsa_circPCMTD1_004 and hsa_circSPECC1_006 expression levels were significantly up-regulated and the expression level of hsa_circZFAND6_008 was significantly down-regulated, consistent with the transcriptome sequencing results (Fig. 4a) (P < 0.01, P < 0.05 ). The expression levels of hsa_circFBXW7_009 and hsa_circZDHHC21_004 increased significantly, contrary to the transcriptome sequencing results (Fig. 4a) (P < 0.01, P < 0.05). The changes of hsa_circUHRF2_003 and hsa_circTCONS_00017881_003 expression were no statistical significance (Fig. 4a). At 24h after 2 Gy irradiation, the expression levels of hsa_circPCMTD1_004 and hsa_circSPECC1_006 were still increased, which were consistent with the transcriptome sequencing results (Fig. 4b) (P < 0.01). The relative expression of hsa_circZFAND6_008, hsa_circFBXW7_009, and hsa_circZDHHC21_004 were also increased, which was contrary to the transcriptome sequencing results (Fig. 4b) (P < 0.01). Hsa_circUHRF2_003 and hsa_circTCONS_00017881_003 expressions were still unchanged.
At 4 h and 24 h post 5 Gy irradiation, hsa_circFAM13B_024, hsa_circMPP6_020, and hsa_circATP5C1_006 expression levels were up-regulated, coincided with the transcriptome sequencing results (Fig. 4c, 4d) (P < 0.01). The expression levels of hsa_circPLOD2_001, hsa_circZFAND6_008, and hsa_circXPO1_021 were decreased significantly, which were reverse to the transcriptome sequencing results (Fig. 4c, 4d) (P < 0.01). The hsa_circBARD1_018 expression level remained unchanged at 4h and increased significantly at 24h (Fig. 4c, 4d) (P < 0.01). The hsa_circNFATC3_003 expressions were no statistically different from those of the control both at 4 h and 24 h after exposure (Fig. 4c, 4d). The above results indicated that the expression levels of 11 out of 14 candidate circRNAs were affected by ionizing radiation, although some of them showed variable trends in contrast to the transcriptome sequencing results. Therefore, these 11 circRNAs have been used for subsequent studies.
Dose-response relationships between ionizing radiation dose and circRNA expressions levels. To test whether the above radiation DE-circRNAs have dose-response effects, the expression levels of the 11 circRNA candidates were further detected in AHH-1 cells at 4 h, 12 h, 24 h, and 48 h after exposed to 0, 2, 4, 6, and 8 Gy 60Co γ -rays. The dose-response relationship between the absorbed doses and the single circRNA expression level was analysed by linear regression analysis. At 4 h after irradiation, the changes of expression levels of 7 circRNAs were overall up-regulated with absorbed doses (Fig. 5b, 5c, 5e-5h, 5k). Of them, the expression of hsa_circFBXW7_009, hsa_circFAM13B_024, and hsa_circPLOD2_001 varied with absorbed doses according to a linear model (Supplementary Table S2). There were no linear relationships between three circRNA expression levels and the absorbed doses, which their expression levels decreased firstly and then increased with the absorbed doses at 4 h post-irradiation (Fig. 5a, 5i, 5j) (Supplementary Table S2). The expression level of hsa_circZFAND6_008 was down-regulated with absorbed doses in a linear model at 4 h after irradiation (Fig. 5d) (Supplementary Table S2). There were also 7 circRNAs whose expression levels were overall up-regulated with absorbed doses at 12 h after irradiation (Fig. 5a-5d, 5f, 5i, 5k). Of them, the expression levels of hsa_circZFAND6_008, hsa_circMPP6_020, and hsa_circPLOD2_001 changed with the absorbed doses following a linear model (Supplementary Table S2). The expression levels of hsa_circZDHHC21_004, hsa_circFAM13B_024, hsa_circATP5C1_006 and hsa_circXPO1_021 were not consistently increased with absorbed doses at 12 h after exposure (Fig. 5e, 5g-5h, 5j). At 24 h after irradiation, 8 circRNAs expression levels gradually increased with absorbed doses followed a linear model (Fig. 5a-5d, 5f-5g, 5i-5j)(Supplementary Table S2). The expression level of Hsa_circZDHHC21_004 was also not stably increased with the absorbed doses at 24 h after radiation exposure (Fig. 5e). The change trend of hsa_circATP5C1_006 expression with the absorbed doses was not significant at 24 h post-irradiation (Fig. 5h). The expression of hsa_circPLOD2_001 was gradually up-regulated from 0 Gy to 6 Gy, but down-regulated at 8 Gy at 24 h after radiation exposure (Fig. 5k). At 48 h after irradiation, the expressions levels of hsa_circFBXW7_009, hsa_circFAM13B_024, and hsa_circPLOD2_001 were significantly increased with absorbed doses in a linear model (Fig. 5c, 5g, 5k) (Supplementary Table S2). The expression of hsa_circBARD1_018 was up-regulated induced by ionizing radiation, but not dose dependent (Fig. 5f). The changed trends of other circRNAs expressions were not stable, some of them were up-regulated firstly but down-regulated later (Fig. 5b, 5d, 5i, 5j) others were fluctuated with the absorbed doses (Fig. 5a, 5e, 5h).
It may be not accurate for dose-estimation by using a single circRNA. Therefore, the expression dosimetry models with multiple circRNAs were established at different time points after ionizing radiation by using of stepwise regression analysis (Table 1). At 4 h after radiation exposure, only the hsa_circPLOD2_001 expression level was included in the model (P < 0.01). Hsa_circPLOD2_001 and hsa_circZFAND6_008 were included in the expression model at 12 h post-irradiation (P < 0.01). The adjusted R2 value of 12 h expression model was greater than that of each single circRNA at the same time point. Collinearity diagnostics analysis showed no collinearity between variables (VIF < 10). The expression model of 24 h after irradiation, hsa_circPCMTD1_004 and hsa_circZFAND6_008 were included (P < 0.01). The adjusted R2 value of 24 h expression model was also greater than that of each single circRNA. Collinearity diagnostic analysis also showed no collinearity between variables (VIF < 10). Only hsa_circPLOD2_001 was included in the expression model at 48 h after irradiation.
Table 1
The expression models established between circRNA expression levels and absorbed doses.
Time-point post-irradiated
|
The circRNAs included in the models
|
Linear regression curvesa
|
Adjusted R2
|
P
|
VIFb
|
4 h
|
Hsa_circPLOD2_001 (x)
|
y = 2.060x-1.990
|
0.950
|
0.003
|
1.000
|
12 h
|
Hsa_circPLOD2_001 (x1)
Hsa_circZFAND6_008 (x2)
|
y = 2.079x1 + 2.618x2-4.820
|
0.996
|
0.002
|
4.472
|
24 h
|
Hsa_circPCMTD1_004 (x1)
Hsa_circZFAND6_008 (x2)
|
y = 6.902x1 + 3.252x2-10.039
|
0.998
|
0.001
|
6.808
|
48 h
|
Hsa_circPLOD2_001 (x)
|
y = 2.200x-1.923
|
0.986
|
0.002
|
1.000
|
a All expression models were established by stepwise regression analysis in this study. y is represented of absorbed doses. x is represented of the normalized expression level of circRNAs. |
b VIF is variance inflation factor. |
The accuracy of above expression models were validated in blind tests (Supplementary Table S3). Although all the adjusted R2 of the each expression model was more than 0.900, most of the estimated doses by using of expression models were deviated from the actual absorbed doses and the relative deviation were higher than 20%. Only the estimated dose in the 3 Gy group at 48 h post-irradiation were similarly to the actual dose, which the relative deviation was less than 10%.