Introduction
Hepatitis B virus (HBV) infection is a serious health problem worldwide. More than 250 million people are chronic hepatitis B (CHB) patients with the potential to cause persistent infection and eventually develop into severe outcome such as liver failure and hepatocellular carcinoma (HCC).1 Interferon alpha (IFNα) or pegylated-interferon alpha (PegIFNα) performs as immune modulators, and nucleos(t)ide analogues (NUCs) with an antiviral role, are currently the main antiviral treatments of HBV.2 PegIFNα can achieve sustained off-treatment control, although it benefits only a fraction of CHB patients and has undesirable side effects.3 NUCs are efficient at controlling HBV replication, meanwhile they should be taken on a long-term for the majority of CHB patients.4 However, hepatitis B surface antigen (HBsAg) seroclearance can be observed in only a tiny portion of CHB patients following long-term treatment.4 Therefore, it is imperative to identify determinants of CHB treatment outcomes to improve personalized therapy. According to several publications, numerous host-related and virus-related factors with an impact on therapeutic efficiency that were deemed outcome predictors for treatment response were ineffective at predicting therapy response at the customized level.5–7
A growing body of research suggests that host genetics may influence CHB patients’ responsiveness to IFNα treatment.8–10 Our 2016 study demonstrated that STAT4_rs7574865 was significantly correlated with PegIFNα therapy response in 466 hepatitis B e antigen (HBeAg)-positive CHB patients.8 Moreover, STAT4_rs7574865 was recently proven to be related to PegIFNα treatment outcomes in CHB patients.9 In 2020, we discovered another single-nucleotide polymorphism (SNP) named CFB_rs12614, which is a susceptibility locus for CHB,11 also had a close association with PegIFNα treatment response.10
As a complement system regulator, CFB plays a vital function in enhancing the complement response and effectively eliminating pathogens.12 Inspired by the study of CFB_rs12614, other potentially functional SNPs in the genes encoding complement regulatory proteins should be investigated to see whether they might predict CHB patients’ abilities in PegIFNα or NUCs treatment response.
Complement regulatory soluble proteins such as C1 inhibitor (C1INH), C4b binding protein (C4BP) and complement factors B, D, H and I (CFB, CFD, CFH and CFI), are more specific, and control either the alternative or the classical or the lectin pathways. Complement regulatory cell-membrane-anchored proteins such as CD35, CD46, CD55, and CD59, control the three complement activation pathways.13
In this study, we comprehensively assessed the efficacy of complement regulators-associated SNPs in predicting the clinical outcomes of CHB treatments by analyzing four trials including patients received PegIFNα treatment or NUCs treatment.
Methods
Patients
A retrospective cohort analysis included 1,763 participants who were recruited in four trials that consisted of EXCEL-PegIFNα cohort 1 (n=186), PB-PegIFNα cohort 2 (n= 708), EFFORT-NUCs cohort 1 (n=553) and EXPLORE–NUCs cohort 2 (n=316). Eligible patients (18 to 65 years of age) enrolled in all four trials had similar inclusion and exclusion criteria, including HBsAg positive for at least 6 months, HBeAg positive, HBV DNA >105 copies/mL, alanine aminotransferase (ALT) ≥ 2× the upper limit of normal (ULN) but <10 ×ULN and without antiviral treatment within 6 months. Other major exclusion criteria included coinfection with hepatitis C, hepatitis D, or the human immunodeficiency virus, presence of chronic liver disease other than viral hepatitis, evidence of hepatic decompensation, portal hypertension, or HCC.
In EXCEL-PegIFNα cohort 1, patients were enrolled across 11 clinical centers in China between April 2010 and October 2010, and received PegIFNα (180 µg/week) for 24 weeks prior to randomization. The levels of serum HBV DNA and HBsAg were detected at week 24. Patients with HBsAg <500 IU/mL and HBV DNA <105 copies/mL were defined as early responders, while others were defined as non-early responders.14 Early responders were treated with PegIFNα for a further 24 weeks (total 48 weeks). Non-early responders continued for another 24 or 72 weeks (total 48 or 96 weeks). PB-PegIFNα cohort 2 involved 43 study locations lasting from January 2013 to August 2015. Eligible patients were randomized to PegIFNα-2b group or PegIFNα-2a group.15 All patients were treated with PegIFNα-2b or PegIFNα-2a at 180 µg/week for 48 weeks and were followed up for 24 weeks. EFFORT-NUCs cohort 1 was conducted at 24 centers in China from August 2009 to March 2012. Patients with suboptimal response (24-week ≥ HBV DNA 300 copies/mL) started telbivudine 600 mg daily and added adefovir (ADV) 10 mg daily from week 28 to week 104, while others received telbivudine monotherapy for 104 weeks.16 Another NUCs cohort was a randomized, and controlled study in 24 centers of China from March 2010 to February 2013. Patients were treated with lamivudine (LAM) for 24 weeks, after which those suboptimal responders with HBV DNA >1,000 copies/mL at week 24 received LAM plus adefovir until week 104, while the early virological responders continued LAM monotherapy.17 In addition, some patients received LAM plus ADV combination therapy for 104 weeks.17 Detailed patient allocation and treatment regimens for each trail have already been described previously.14–17 Additionally, each trial’s demographic, virologic, and clinical parameters have been reported in previous.9 The study protocol followed the 2013 revision of the guidelines of the Declaration of Helsinki and the Nanfang Hospital Ethics Committee authorized this study. All patients signed an informed consent form.
Serological testing
Throughout the period of four trials, clinical, and laboratory evaluations were done every 8–16 weeks. The Roche COBAS Taqman platform and ARCHITECT i2000SR (Abbott Laboratories, Chicago, IL, USA) were utilized to detect the levels of HBV DNA and other virological indicators. The results were replaced with a lower limit of detection (LLOD) of 12 IU/mL or 69.84 copies/mL when they were lower than the LLOD.
Efficacy measures
The combined response (CR) rate and the level of HBsAg loss were considered as endpoint responses to assess therapy efficacy. The combination of HBeAg seroconversion (removal of HBeAg and the presence of HBeAg antibody) and reduced HBV DNA level (HBV DNA <2,000 IU/mL at week 72 for the PegIFNα cohorts or HBV DNA <12 IU/mL at week 104 for the NUCs cohorts) was classified as CR. The level of HBsAg in serum less than 0.05 IU/mL at week 72 in the PegIFNα cohorts or at week 104 in the NUCs cohorts was regarded as HBsAg loss. The detailed information was described previously.14–17
SNP selection and genotyping
The complement regulators-associated SNPs were selected through a screening process shown in Supplementary Figure 1. Briefly, the genes coding complement regulatory proteins were annotated by GENCODE, Build 37 (Supplementary Table 1). Based on the 1000 Genomes Project database, 512 SNPs located in the genes coding complement regulatory proteins were included after quality control (minor allele frequency >0.05, missing rate <0.05, p-value of Hardy-Weinberg equilibrium >0.05). To focus on potentially functional SNPs that may affect gene expression, an expression quantitative trait locus (eQTL) analysis in whole blood was conducted (SNPs with p < 0.01 were considered significant), and 103 SNPs were remained (Supplementary Table 2). As some of the 103 SNPs are in high linkage disequilibrium (LD) with each other, a clump analysis was performed to screen out independent SNPs. Nine independent SNPs were chosen for further investigation (Supplementary Table 3). Peripheral blood was used to extract genomic DNA with a QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany) as per the manufacturer’s instructions. Genotyping of the nine selected SNPs in genomic DNA samples of all the CHB patients was analyzed by the Sequenom mass array platform (Sequenom, San Diego, CA, USA).
Statistical analysis
The chi-squared test or Fisher exact test for categorical parameters and the Student t-test or the Mann–Whitney test for continuous parameters were used to evaluate qualitative and quantitative differences across subgroups, as appropriate. By computing the frequency of favorable alleles, a polygenic score (PGS) was established to measure the cumulative effects of several SNPs recognized as indicators of CHB treatment response. Patients who possessed more favorable alleles had a higher level of CR or HBsAg loss. The Cochran-Armitage trend test was performed to analyze the correlations of PGS with CR level and HBsAg loss. All statistical tests were two sided. Two-tailed p-values <0.05 were considered statistically significant. SAS 9.3 (SAS Institute, Cary, NC, USA) was used to perform the statistical analysis.
Results
Patient characteristics
The correlations of complement regulators-associated SNPs with CHB therapies were evaluated in two PegIFNα cohorts with total 894 CHB patients and two NUCs cohorts with total 869 CHB patients. Patients in the PegIFNα and NUC cohorts were comparable in baseline demographic, virological, and clinical characteristics, as shown in Table 1. Male patients accounted for 74.16% of patients received PegIFNα treatment and 79.29% of patients received NUCs treatment. The majority of patients were Han Chinese (94.18% in PegIFNα-treated patients and 96.78% in NUCs -treated patients) and only Han Chinese were involved in the subsequent analyses for reducing genetic heterogeneity. Most of the CHB patients infected with genotype B or C HBV, which took over approximately 98.00% of all the CHB patients. The baseline levels of HBV serological markers in PegIFNα-treated patients were almost the same as the NUCs-treated patients. Moreover, the level of ALT was higher in PegIFNα-treated patients than in NUCs-treated patients.
Table 1Characteristics of the patients treated with PegIFNα and NUCs
| Baseline variable | PegIFNα-treated patients
| NUC-treated patients
| p-value (PegIFNα cohort 1+2 vs. NUC cohort 1+2) |
|---|
| PegIFNα cohort 1 (n=186) | PegIFNα cohort 2 (n=708) | PegIFNα cohort 1+2 (n=894) | NUC cohort 1 (n=553) | NUC cohort 2 (n=316) | NUC cohort 1+2 (n=869) |
|---|
| Male sex (%) | 151 (81.18) | 512 (72.30) | 663 (74.16) | 450 (81.40) | 239 (75.60) | 689 (79.29) | 0.01 |
| Age, years; mean (SD) | 28.44 (6.51) | 29.74 (6.70) | 29.47 (6.68) | 30.13 (8.96) | 31.94 (9.40) | 30.79 (9.16) | 5.93×10−4 |
| Han ethnicity (%) | 184 (98.92) | 658 (92.90) | 842 (94.18) | 540 (97.70) | 301 (95.30) | 841 (96.78) | 0.53 |
| HBV genotype (%) | | | | | | | 0.03 |
| B | 68 (36.56) | 278 (40.40) | 346 (38.70) | 215 (38.90) | 115 (36.40) | 330 (37.97) | |
| C | 116 (62.37) | 403 (58.50) | 519 (58.05) | 335 (60.60) | 199 (63.00) | 534 (61.45) | |
| Others | 2 (1.08) | 8 (1.20) | 10 (1.11) | 3 (0.50) | 2 (0.60) | 5 (0.59) | |
| NA | 0 (0.00) | 19 (2.70) | 19 (2.13) | 0 (0.00) | 0 (0.00) | 0 (0.00) | |
| HBV DNA#, log10IU/mL; mean (SD) | 7.64 (1.29) | 7.93 (0.76) | 7.87 (0.90) | 7.71 (1.07) | 7.80 (0.90) | 7.74 (1.01) | 5.10×10−3 |
| HBsAg#, log10IU/mL; mean (SD) | 3.96 (0.77) | 4.28 (0.53) | 4.21 (0.60) | 4.20 (0.68) | 4.18 (0.72) | 4.19 (0.69) | 0.49 |
| HBeAg#, log10PEIU/mL; mean (SD) | 2.42 (0.98) | 3.04 (0.56) | 2.91 (0.72) | 2.74 (0.64) | 2.72 (0.62) | 2.73 (0.63) | 3.52×10−8 |
| ALT#, × ULN; mean (SD) | 4.58 (4.02) | 4.54 (2.13) | 4.55 (2.65) | 4.30 (3.79) | 3.40 (2.52) | 3.98 (3.41) | 1.05×10−4 |
| STAT4_rs7574865 genotype (%) | | | | | | | 0.89 |
| GG | 98 (52.69) | 329 (46.47) | 427 (47.76) | 271 (49.01) | 142 (44.94) | 413 (47.50) | |
| GT | 71 (38.17) | 303 (42.80) | 374 (41.83) | 223 (40.33) | 142 (44.94) | 365 (42.00) | |
| TT | 17 (9.14) | 65 (9.18) | 82 (9.17) | 55 (9.95) | 31 (9.81) | 86 (9.90) | |
| NA | 0 (0.00) | 11 (1.55) | 11 (1.23) | 4 (0.72) | 1 (0.32) | 5 (0.60) | |
| CFB_rs12614 genotype (%) | | | | | | | 0.85 |
| CC | 172 (92.47) | 651 (91.95) | 823 (92.06) | 508 (91.86) | 294 (93.04) | 802 (92.29) | |
| CT | 14 (7.53) | 44 (6.21) | 58 (6.49) | 38 (6.87) | 22 (6.96) | 60 (6.90) | |
| TT | 0 (0.00) | 2 (0.28) | 2 (0.22) | 3 (0.54) | 0 (0.00) | 3 (0.35) | |
| NA | 0 (0.00) | 11 (1.55) | 11 (1.23) | 4 (0.72) | 0 (0.00) | 4 (0.46) | |
| CD55_rs28371597 genotype (%) | | | | | | | 0.55 |
| GG | 171 (91.94) | 598 (84.46) | 769 (86.02) | 491 (88.79) | 278 (87.97) | 769 (88.49) | |
| GT | 12 (6.45) | 48 (6.78) | 60 (6.71) | 42 (7.59) | 22 (6.96) | 64 (7.36) | |
| TT | 1 (0.54) | 1 (0.14) | 2 (0.22) | 1 (0.18) | 1 (0.32) | 2 (0.23) | |
| NA | 2 (1.08) | 61 (8.62) | 63 (7.05) | 19 (3.43) | 15 (4.75) | 34 (3.91) | |
Additionally, as mentioned earlier, our group indicated a statistically significant connection between STAT4_rs7574865 and CFB_rs12614 and PegIFNα treatment response.9,10 We found that the distributions of the genotypes of these two SNPs were very similar in PegIFNα-treated patients and NUCs-treated patients. Except for the above two SNPs, the genotype distributions of rs28371597 (including GG, GT, and TT) were roughly the same between patients received PegIFNα or NUCs therapies.
Complement regulator-associated SNPs were selected via the screening procedure
Through the screening process as shown in Supplementary Figure 1, nine SNPs were screened out from the complement regulatory genes, such as CD55, CABPA, CFB, CFD, CR1 and CINH (Supplementary Table 3). The correlations between the alleles of the above nine SNPs and CHB treatments were assessed in the separate four cohorts. Among the nine SNPs, only rs28371597, located in the intronic region of CD55, showed significant correlation with CR in both the two PegIFNα cohorts (PegIFNα cohort 1: p=0.01, OR=3.72; PegIFNα cohort 2: p=8.57 × 10−3, OR=2.20; Table 2). However, in the separate four cohorts, no other SNPs were statistically correlated with CR in PegIFNα-treated patients. All nine SNPs did not show statistically significant correlations with CR in NUCs-treated patients.
Table 2Associations of the nine complement regulators-associated SNPs with CR in the two PegIFNα cohorts and two NUCs cohorts
| SNP | Allele# | PegIFNα cohort 1
| PegIFNα cohort 2
| NUCs cohort 1
| NUCs cohort 2
|
|---|
| MAF | OR (95% CI) | p | MAF | OR (95% CI) | p | MAF | OR (95% CI) | p | MAF | OR (95% CI) | p |
|---|
| rs12057769 | A/G | 0.16 | 1.17 (0.63–2.19) | 0.61 | 0.20 | 1.06 (0.76–1.48) | 0.74 | 0.19 | 1.14 (0.80–1.61) | 0.47 | 0.20 | 1.20 (0.73–2.00) | 0.47 |
| rs4266889 | A/G | 0.27 | 1.12 (0.68–1.87) | 0.65 | 0.31 | 0.92 (0.68–1.24) | 0.57 | 0.27 | 0.95 (0.70–1.31) | 0.77 | 0.32 | 1.10 (0.71–1.72) | 0.67 |
| rs28371597 | T/G | 0.04 | 3.72 (1.2 6–10.99) | 0.01 | 0.04 | 2.20 (1.21–4.02) | 8.57×10−3 | 0.04 | 0.88 (0.43–1.81) | 0.73 | 0.04 | 0.66 (0.19–2.25) | 0.50 |
| rs6662070 | G/A | 0.40 | 1.11 (0.70–1.77) | 0.65 | 0.43 | 1.00 (0.76–1.32) | 1.00 | 0.41 | 1.04 (0.78–1.37) | 0.80 | 0.44 | 0.94 (0.62–1.44) | 0.78 |
| rs1507765 | A/C | 0.41 | 0.98 (0.62–1.56) | 0.93 | 0.39 | 1.22 (0.93–1.61) | 0.16 | 0.41 | 0.89 (0.67–1.18) | 0.41 | 0.39 | 0.99 (0.64–1.52) | 0.96 |
| rs1048709 | A/G | 0.31 | 1.18 (0.72–1.91) | 0.52 | 0.33 | 0.89 (0.66–1.19) | 0.42 | 0.34 | 0.90 (0.67–1.21) | 0.48 | 0.29 | 1.04 (0.66–1.65) | 0.87 |
| rs541862 | C/T | 0.06 | 1.16 (0.46–2.91) | 0.75 | 0.05 | 1.25 (0.69–2.26) | 0.46 | 0.05 | 0.90 (0.48–1.71) | 0.75 | 0.06 | 1.44 (0.66–3.13) | 0.35 |
| rs78364821 | T/C | 0.16 | 0.83 (0.43–1.60) | 0.58 | 0.16 | 1.01 (0.69–1.48) | 0.95 | 0.14 | 1.01 (0.67–1.50) | 0.98 | 0.17 | 1.29 (0.75–2.19) | 0.36 |
| rs3826945 | C/T | 0.41 | 1.43 (0.84–2.44) | 0.19 | 0.38 | 1.25 (0.90–1.73) | 0.18 | 0.38 | 0.90 (0.64–1.27) | 0.55 | 0.40 | 1.03 (0.63–1.70) | 0.90 |
Rs28371597 was correlated with the levels of CR and HBsAg loss in CHB patients received PegIFNα treatment
Rs28371597 possesses a G major allele and a T minor allele. Because the T allele of rs28371597 is uncommon, the GT and TT genotypes were combined into the GT/TT genotype to proceed with the succeeding investigations. PegIFNα-treated patients with the rs28371597_GG genotype were found to have significantly lower CR rate compared to those with the rs28371597_GT/TT genotype (PegIFNα cohort 1: 25.73% vs. 53.85%, p=0.03; PegIFNα cohort 2: 18.73% vs. 32.65%, p=0.02; PegIFNα cohort 1 plus cohort 2: 20.29% vs. 37.10%, p=2.00 × 10−3; Fig. 1). When other basic factors, such as HBV genotype, HBV DNA level, rs7574865, and rs12614 were modulated, the correlation remained statistically significant (p=0.04, 0.04 and 0.02 for the PegIFNα cohort 1, 2, and their combination, respectively, Table 3 and Supplementary Table 4).
Table 3Multivariate logistic regression analysis of rs28371597 with CR in the two PegIFNα cohorts
| PegIFNα cohort 1
| PegIFNα cohort 2
| PegIFNα cohort 1+2
|
|---|
| OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p |
|---|
| Trial (PegIFNα cohort 1 vs. 2) | | | | | | |
| Sex (Female vs. Male) | | | | | | |
| Age, years | | | 0.94 (0.91–0.98) | 1.75×10−3 | 0.96 (0.93–0.99) | 4.68×10−3 |
| HBV genotype (B vs. C) | | | 1.60 (1.06–2.40) | 0.03 | 1.61 (1.12–2.33) | 0.01 |
| HBV DNA#, log10 IU/mL | | | | | | |
| HBsAg#, log10 IU/mL | | | | | 0.74 (0.52–1.04) | 0.08 |
| HBeAg#, log10 PEIU/mL | 0.47 (0.32–0.69) | 1.04×10−4 | 0.50 (0.35–0.72) | 1.61×10−4 | 0.57 (0.43–0.76) | 9.80×10−5 |
| ALT#, loge ×ULN | 2.49 (1.42–4.36) | 1.40×10−3 | | | 1.83 (1.29–2.58) | 6.08×10−4 |
| rs7574865 genotype | 1.85 (1.09–3.12) | 0.02 | | | 1.39 (1.07–1.80) | 0.01 |
| rs12614 genotype | | | 4.19 (0.98–17.97) | 0.05 | 3.85 (1.15–12.82) | 0.03 |
| rs28371597 genotype | 3.52 (1.03–12.00) | 0.04 | 2.01 (1.03–3.91) | 0.04 | 1.99 (1.11–3.56) | 0.02 |
We also examined the levels of HBeAg seroconversion, HBV DNA and CR both in patients who carried the rs28371597_GG genotype and in patients who carried the rs28371597_GT/TT genotype for the given time periods. The seroconversion rate of HBeAg was always greater in rs28371597_GT/TT, and the average levels of HBV DNAs in patients with rs28371597_GT/TT were consistently lower since 24 weeks (Supplementary Fig. 2A–D). As for the CR rate, it was lower in patients with the rs28371597_GG genotype than that in patients with the rs28371597_GT/TT genotype since 48 weeks in the PegIFNα cohort 1. While in the PegIFNα cohort 2, it was always lower in the rs28371597_GG genotype carriers than in the rs28371597_GT/TT genotype carriers (Supplementary Fig. 2E–F). HBsAg loss was also compared between patients with different rs28371597 genotypes. The levels of HBsAg loss in patients that possessed the rs28371597_GG genotype were persistently lower than in patients that possessed the GT/TT genotype (PegIFNα cohort 1: 2.34% vs. 7.69%; PegIFNα cohort 2: 2.34% vs. 4.08%; PegIFNα cohort 1 plus cohort 2: 2.34% vs. 4.84%), although the difference did not reach statistical significance owing to the limited number of patients with HBsAg loss (Fig. 1D–F).
PGS was correlated with the levels of CR and HBsAg loss in CHB patients received PegIFNα treatment
Recently, our group used CFB_rs12614 and STAT4_rs7574865 to establish the PGS, and discovered that PGS became larger, the CR rate gradually increased.10 Here, we tried to add CD55_rs28371597 into the previous PGS, aiming to improving the ability of predicting individual therapy efficiency. The favorable alleles of rs28371597, rs12614, and rs7574865 were T, C, and T, respectively. The number of the favorable alleles that the CHB patients carried reflected the level of PGS. As expected, the higher level of PGS, the higher CR rate was obtained (Fig. 2A–C). For example, we observed that the CR rates in all PegIFNα-treated patients who carrying 0, 1, 2, 3, 4, and 5 favorable alleles corresponded to 0.00%, 3.45%, 18.43%, 22.12%, 34.04% and 62.50%, respectively (p-trend=7.94×10−6; Fig. 2C). PGS was significantly associated with CR even after the adjustment for other basic factors (Table 4 and Supplementary Table 5).
Table 4Multivariate logistic regression analysis of PGS with CR in the two PegIFNα cohorts
| Parameters | PegIFNα cohort 1
| PegIFNα cohort 2
| PegIFNα cohort 1+2
|
|---|
| OR (95% CI) | p | OR (95% CI) | p | OR (95% CI) | p |
|---|
| Trial (PegIFNα cohort 1 vs. 2) | – | – | – | – | – | – |
| Sex (Female vs. Male) | – | – | – | – | – | – |
| Age, years | – | – | 0.95 (0.91–0.98) | 2.32×10−3 | 0.96 (0.93–0.99) | 3.88×10−3 |
| HBV genotype (B vs. C) | – | – | 1.63 (1.08–2.44) | 0.02 | 1.66 (1.15–2.40) | 6.57×10−3 |
| HBV DNA#, log10 IU/mL | – | – | – | – | – | – |
| HBsAg#, log10 IU/mL | – | – | – | – | 0.74 (0.52–1.04) | 0.08 |
| HBeAg#, log10 PEIU/mL | 0.48 (0.33–0.71) | 2.05×10−4 | 0.48 (0.33–0.69) | 7.74×10−5 | 0.56 (0.43–0.75) | 6.49×10−5 |
| ALT#, loge ×ULN | 2.59 (1.48–4.55) | 9.20×10−4 | – | – | 1.87 (1.32–2.63) | 3.65×10−4 |
| PGS | 2.20 (1.39–3.50) | 7.84×10−4 | 1.47 (1.13–1.91) | 4.21×10−3 | 1.56 (1.24–1.95) | 1.27×10−4 |
The association analysis of PGS with the level of HBsAg loss yielded similar results (Fig. 2D–F). The levels of HBsAg loss were rising with carrying increased favorable alleles in patients received PegIFNα treatment (favorable alleles with 0, 1, 2, 3, 4, and 5, corresponding to 0.00%, 0.00%, 1.36%, 3.33%, 4.26% and 12.50%, respectively; p-trend=9.40×10−3; Fig. 2F). Even after the adjustment for other basic factors, the correlations of PGS with HBsAg loss still reached a statistically significant level in PegIFNα-treated patients (Supplementary Tables 6 and 7).
Discussion
In this study, we examined the correlations of complement regulators-associated SNPs with responses to CHB treatments in four independent, retrospective, multicenter, well-characterized cohorts. We demonstrated that rs28371597 located in CD55 intron was strongly correlated with the level of CR in patients received PegIFNα treatment but not NUCs treatment. The new PGS model composed of CFB_rs12614, STAT4_rs7574865, and CD55_rs28371597, was strongly linked with the levels of CR and HBsAg loss and predicted a better off-treatment outcome individually in PegIFNα-treated patients.
The complement system is a key element of the innate immune response and can be triggered through the classical, lectin, or the alternative pathway.18 All pathways converge to produce protein complexes (C3 convertases) which cleave and activate the core C3 component into C3a and C3b which is a component of the C5 convertase. C5 divides C5 into C5a and C5b, the latter forms a MAC with C6-C9 at last.19 When activated, the components of the complement system are crucial to fight pathogens, target invaders for destruction, and initiate inflammatory responses.20 Regulation of the complement system is managed by C1INH, C4BP, complement factors B, D, H and I (CFB, CFD, CFH and CFI), CD35, CD46, CD55, and CD59.13 These molecules are either co-factors (CFH, C4BP, CD35 and CD46) in the CFI-mediated cleavage of C3b (and/or C4b) or can promotes the disassembly (CD35 and CD55) of the existing C3 convertases.21,22 In addition, CFB and CFD both regulate the formation of C3 convertases.12 These complement regulators work to guarantee that the rate and extent of complement activation are proportionate to the duration and strength of the activating signal.
Besides the importance as a regulator of the complement system, complement regulatory proteins modulate T cell function and survival. Some studies have reported that complement regulatory proteins has a role in the interaction between the complement system and T cell immunity, which had been widely considered as a crucial player in HBV clearance.23–25 For example, activation by CD46, a membrane-bound protein of the complement system, contributed to a strong proliferative response of activated T cells,23 and induced IFNγ secretion for differentiation into Th1 cells.24 The immunomodulatory role of CD59 in T cells has also been reported.25 Our previous study indicated there was significant correlation of CFB_rs12614 with PegIFNα therapy response in HBeAg-positive CHB patients,10 which prompted us to explore the correlations of complement regulators-associated SNPs with CHB treatments in CHB patients. Only CD55_rs28371597, one of nine SNPs chosen from genes encoding for complement regulatory proteins, was shown to be strongly related to PegIFNα therapy response in patients with CHB.
CD55, also known as decay accelerating factor (DAF), is a cell surface molecule that is glycosylphosphatidylinositol (GPI)-anchored. CD55 is not only a complement inhibitor, but also an immune modulator. Liu et al. showed the mouse lacking the Daf1 gene could significantly promotes T cell responses with the increased IFN-γ and decreased IL-10 production.26 Other investigators have found that down-regulation of CD55 in T cells enhances T cell proliferation, suggesting its close association with T cell responsiveness.27 A study by Capasso et al. discovered that CD55 combined to its ligand CD97 and then induced T cell strong response involving increased T cell proliferation and up-regulation of activation markers CD69 and CD25.28 Furthermore, CD55 was also found to inhibit natural killer (NK) cells.29 The findings show that both innate and adaptive immune responses are regulated by CD55.
Except for the direct antiviral effect, IFNα has another effect of immune modulation on both innate and adaptive immune response.30 It is likely that CD55 links to IFNα treatment efficacy in an immunoregulatory manner. One possibility is that CD55 has influence on IFNα treatment via modulating T cell-mediated immunity, supported by the fact that IFNα regulates the T cell response which is important in establishing a functional cure of chronic HBV infection.31 Additionally, the effect of CD55 on IFNα treatment efficacy through regulation of NK cells seems rational and might be explained by the fact that HBV suppression during IFNα treatment coincides with a rise in the number of circulating CD56bright NK cells, increased expression of the cytotoxic receptor TNF-related apoptosis-inducing ligand (TRAIL) by NK cells.32,33 Moreover, during the above actions of immune responses, CD55-mediated IFNγ production may also contribute to durable IFNα treatment response.32,33 Taken together, CD55 may play an integral part in achieving the control of HBV infection with IFNα treatment response by regulating T cell immunity, NK cells activation and IFNγ production.
As mentioned above, rs28371597 is located in the intron region of CD55. Other than genetic polymorphisms in the coding region causing missense mutations which result in protein dysfunction, intronic polymorphisms may locate in the regulatory element sites, such as cis-regulatory element region, suggesting that they may have a biological role by regulating gene expression.34 HaploReg v4.1 database shows that the regulatory motifs of several well-known transcription factors including DMRT4 and Foxp3, change because of the different alleles of rs28371597, thereby affecting their binding activities. In addition, rs28371597 may be just a tagging SNP, which is in high LD with the causal SNP. One assumption is that rs28371597 or another causal SNP which is in high LD with rs28371597 affects CD55 mRNA stability or protein synthesis via allele-imbalanced regulatory proteins, thus influencing its role for host immunity and then the response to IFNα treatment. Based on these studies, we hypothesized that the transcription factors DMRT4 or Foxp3 were less bound to the favorable allele T of rs28371597 and resulted in decreased expression of CD55. Subsequently, downregulation of CD55 promoted T cell response and the number of circulating CD56bright NK cells because of high IFNγ production and TRAIL expression. Finally, the T allele of rs28371597 had a role in the effectiveness of IFNα treatment (Fig. 3). Nevertheless, additional study is needed to discover how the CD55 intronic SNP influences IFNα treatment efficiency.
The available evidence indicates that PGS has both dependable and clinically significant effectiveness.10 We previously constructed a PGS with only two SNPs to assess the combined impact of CFB_rs12614 with STAT4_rs7574865, and found it effectively reflected clinical outcomes of CHB patients received PegIFNα treatment.10 We anticipate that when additional pharmaco-related SNPs are discovered, the PGS including more SNPs will be increasingly successful in predicting PegIFNα treatment response. In this study, we added CD55_rs28371597 into the previous PGS and constructed a new PGS model. The level of CR in PegIFNα-treated patients with the PGS model scored 5 was much higher than in patients with the PGS model scored less than 1 (62.5% vs. 3.45%). The level of HBsAg loss in PegIFNα-treated patients with the PGS model scored 5 was also much higher than in patients with the PGS model scored less than 1 (12.5% vs. 0.00%). Overall, a PGS can improve the ability of predicting individual therapy efficiency.
The differential performance of CD55_rs28371597 for predicting CHB therapy response to PegIFNα vs. NUCs is most probably owing to their distinct functional mechanisms. NUCs therapy suppresses HBV replication by targeting reverse transcriptase of HBV, but not through an immunoregulatory effect.4 The major effect of PegIFNα is immune response that targets HBV-infected hepatocytes, thus decreasing the number of cells containing the HBV cccDNA.35 PegIFNα is a potential therapeutic agent with the advantage of achieving off-treatment outcomes probably because of the specific and efficient mechanism involving both immunomodulation and viral suppression.
It should be emphasized that this study has numerous limitations. First, the association of CD55_rs28371597 with HBsAg loss did not reach a statistically significant level in either the two PegIFNα cohorts or their combination. One explanation is that the relatively small sample of patients with HBsAg loss lacks sufficient statistical power to confirm the connection, and which may require additional validation by other larger investigations in the future. Second, because of the relatively small size of PegIFNα cohort 1, the PGS was statistically linked to HBsAg loss in PegIFNα cohort 2. Third, because the HBV genotypes in our patients were primarily genotype B and genotype C, the efficacy of rs28371597 with other HBV genotypes was not discovered and has to be explored further.