Resistin reinforces interferon λ-3 to eliminate hepatitis C virus with fine-tuning from RETN single-nucleotide polymorphisms

The effect of resistin (RETN) on the response to anti-HCV therapy remains unclear. A prospective cohort study was performed using 655 consecutive HCV patients, of whom 513 had completed a course of interferon-based therapy. Multivariate and GEE analyses revealed four RETN single-nucleotide polymorphisms (SNPs), rs34861192, rs3219175, rs3745367 and rs1423096, to be synergistically associated with resistin levels. After adjusting for co-factors such as interferon λ-3 (IFNL3)-rs12979860, the resistin level and the hyper-resistinemic genotype at the 4 RETN SNPs were positively and negatively associated with a sustained virological response (SVR), respectively. RETN-rs3745367 was in linkage disequilibrium with IFNL3-rs12979860. Compared to non-SVR patients, SVR patients had higher levels of pre-therapy resistin, primarily originating from intrahepatic lymphocytes, stellate cells, Kupffer cells, hepatic progenitor cells and hepatocytes. This difference diminished over the course of therapy, as only SVR patients exhibited a 24-week post-therapy decrease in resistin. Both resistin and IFNL3 mRNAs were upregulated, but only resistin mRNA was upregulated by recombinant resistin in peripheral blood mononuclear cells with and without hyper-resistinemic genotypes of the 4 RETN SNPs, respectively. Fine-tuned by RETN SNPs, intrahepatic, multi-cellular resistin reinforced IFNL3 in eliminating HCV via immunomodulation to counteract pro-inflammation. These results encourage the development of novel resistin-targeted anti-viral agents.


Study design. As shown in Supplementary
, 840 consecutive patients positive for anti-HCV Ab (Abbott Laboratories, IL, USA) for more than 6 months were surveyed for participation in the current study at a tertiary referral center between July 2009 and June 2015. In total, 655 CHC patients were recruited, of whom 513 had completed a course of anti-HCV therapy with pegylated interferon α -2b (1.5 μ g/kg/week) and ribavirin (800-1400 mg/day) for up to 24 or 48 weeks according to response-guided therapeutic criteria 3 . HCV RNA levels and genotypes were tested using COBAS Amplicor and InoLipa methods (Roche Diagnostics, Tokyo, Japan). SNPs of interferon λ -3 (IFNL3) or interleukin-28B-rs12979860 were assessed using genomic DNA, as described previously 3 . Preliminary screening of 10 RETN SNPs, including 4 in the promoter, 1 in the second intron, and 5 in the downstream region of RETN at 19p13.2 (rs1862513, rs10401670, rs34124816, rs3219175, rs3745367, rs3745369, rs1477341, rs4804765, rs1423096 and rs34861192) 14,15 , were assessed in 200 randomly selected patients of the 655 CHC patients using TaqMan SNP Genotyping Assays (Applied Biosystems, MA, USA) (Supplementary Table 1).
The following baseline factors were evaluated for all 655 patients: sex, age and body mass index (BMI); IFNL3-rs12979860 and RETN SNPs, including rs1423096, rs34861192, rs3745367 and rs3219175 genotypes; the HCV RNA level and HCV genotype; aspartate aminotransaminase (AST), alanine aminotransaminase (ALT), fasting glucose and insulin levels; homeostasis model assessment of insulin resistance (HOMA-IR); total cholesterol (TC), triglycerides (TGs), high-sensitivity C-reactive protein (hs-CRP), ferritin, thyroid-stimulating hormone (TSH), creatinine, complement component 3 (C3), and C4 levels; white blood cell (WBC) count and platelet count; the AST-to-platelet ratio index (APRI); the estimated glomerular filtration rate (GFR) calculated using the Cockcroft-Gault equation; and resistin levels (R&D Systems, MN, USA). For the 513 patients who had completed anti-HCV therapy, all the aforementioned factors were evaluated at 2 weeks before therapy, after 4, 12 and 24 weeks of therapy, at the end of therapy, and at 12 and 24 weeks after the end of therapy. IR was defined as a HOMA-IR score ≥2. 5. SVR was defined as undetectable levels of HCV RNA at 24 weeks after therapy completion. All of the serum biochemistry and viral marker measurements were performed using routine automated techniques at the clinical pathology or liver research laboratories of the hospital. Twenty normal male controls were enrolled for PBMC stimulation testing and a fasting peripheral blood (PB) smear. Ten paraffinized normal liver samples were acquired from the tissue bank of the hospital to perform liver histology studies. For immunohistochemical (IHC) analyses, fasting PB smears and liver biopsy samples were acquired from 20 CHC patients with SVR and 20 sex-and age-matched CHC patients without SVR before anti-HCV therapy.
Functional assays between IFNL3 and resistin. PBMCs were isolated from 1 ml of whole blood from 20 normal subjects in triplicate using density gradient centrifugation with BD Vacutainer CPT heparin tubes. One day before the experiment, the PBMCs were plated at a density of 1 × 10 6 cells/well in 24-well plates. After replacing the medium, recombinant resistin, IFNL3 proteins, and lipopolysaccharide (LPS) (R&D Systems, Minneapolis, MN) were added at a concentration of 100 ng/ml. As LPS induces resistin gene expression in macrophages 19 , it was used as a positive control. PBMCs from the same individuals not exposed to recombinant proteins served as the controls. The media were harvested in triplicate at 1, 2 and 4 hr after the recombinant proteins were added. The RNA levels of resistin and IFNL3 were assessed using quantitative reverse transcription real-time polymerase chain reaction (qRT-PCR). The primer sequences used for the qRT-PCR analysis of resistin and IL28B are listed in Supplementary Statistics. All statistical analyses were performed using Statistical Product and Service Solutions software (SPSS package version 21, SPSS Inc., Chicago, USA). To compare different variables in two different groups, continuous variables were analyzed using Student's t-tests; categorical variables were analyzed using chi-squared tests or Fisher's exact tests, as appropriate. Groups of more than 2 were compared by one-way analysis of variance (ANOVA) for different variables, followed by post hoc tests. Univariate and multivariate analyses were used to determine relationships between the dependent and independent variables cross-sectionally, and generalized estimating equation (GEE) repeated measures tests were applied to determine relationships longitudinally. Paired t-tests were used to compare variables within individuals before and at 24 weeks after anti-HCV therapy. In addition to dominance-recessive mode analysis, for all SNPs evaluated, we consistently encoded the major homozygous genotype as 0, the heterozygous genotype as 1, and the minor homozygous genotype as 2 20 . Nonparametric statistics were applied as indicated. Hardy-Weinberg equilibrium and minor allele frequencies of the genotypes were determined to narrow down the 10 RETN SNPs. The statistical significance of the genotype correlations was evaluated using linear-by-linear association. Visualization of the correlation was achieved using generalized association plots 21 . The diagnostic accuracy of using pre-therapy resistin to predict therapeutic response was assessed in terms of true positive (sensitivity) versus true negative (1-specificity) responses using

Results
Four RETN SNPs, including rs34861192, rs3219175, rs3745367 and rs1423096, were synergistically associated with resistin levels. The baseline characteristics of the 655 CHC patients and comparisons between those with and without SVR are listed in Table 1. Among the 513 CHC patients who finished anti-HCV therapy, higher levels of pre-therapy resistin and platelet count, higher percentages of the IFNL3-rs12979860 CC genotype and HCV genotype 2, lower levels of HCV RNA, HOMA-IR and triglycerides, and lower percentages of HCV genotype 1 were found in those with SVR (n = 428) compared with those without SVR (n = 85) ( Table 1).
Among the 10 RETN SNPs (Supplementary Table 1) investigated, the genotype distributions of one SNP (rs1862513) significantly deviated from Hardy-Weinberg equilibrium (p < 0.05) and were not further analyzed. Two SNPs (rs34124816 and rs4804765) have low minor allele frequencies in the Taiwanese population (<15%) and were de-prioritized for further analysis. Of the remaining 4 downstream SNPs, the genotypes correlated with each other (Pearson's correlation > 0.3, p < 0.001). Thus, SNP rs1423096, which is the farthest SNP downstream and which did not overlap with the next gene (MCEMP1), was selected as a representative SNP. Consistently, the genotypes of the excluded 6 RETN SNPs did not correlate with pre-therapy resistin levels. In contrast, the genotypes of the 4 selected SNPs, rs34861192 and rs3219175 in the promoter, rs3745367 in the second intron, and rs1423096 (all p < 0.001) in the downstream region of RETN, correlated highly with pre-therapy resistin levels. As determined by post hoc tests, the "TT" and "TC" variants of rs1423096 (p < 0.001), "AA and GA" variants of rs34861192 (p < 0.001), "AA" variant of rs3745367 (p = 0.01), and "AA" and "GA" variants of rs3219175 (p < 0.001) showed higher pre-therapy resistin levels than the other variants of the same alleles. We stratified the patients into 5 groups according to RETN scores, which ranged from 0 to 4 and were defined by the number of the hyper-resistinemic genotypes among the 4 RETN SNPs (e.g., for rs1423096, hyper-resistinemic genotypes "TT" and "TC" were coded as "1", and the normal resistinemic genotype "CC" was coded as "0") (Supplementary Table 3). The 4 RETN SNPs had synergistic effects on the pre-therapy resistin levels, and RETN scores positively correlated with pre-therapy resistin levels (p < 0.001) (Fig. 1a). Univariate and multivariate analyses showed that the RETN score, pre-therapy ferritin level, WBC count and GFR were independent factors of pre-therapy resistin levels ( Table 2).

The 4 RETN SNPs and SVR affected the longitudinal trend of the resistin level. For the 513 CHC
patients who completed anti-HCV therapy, GEE analysis revealed that the RETN score, SVR, therapeutic intervention, ALT and APRI levels, and WBC count determined the longitudinal trend of the resistin level ( Table 3).
The 4 RETN SNPs and resistin were associated with SVR after adjustment for confounding factors, including the IFNL3 SNP genotype. As shown in Table 4, the IFNL3 SNP, HCV genotype, HCV RNA, pre-therapy TG and platelet levels were factors for SVR in our univariate analysis. BMI, the IFNL3 SNP, HCV genotype, RETN score, and pre-therapy resistin and ferritin levels were independently associated with SVR in multivariate analysis.
Higher pre-therapy resistin levels were observed in CHC patients with SVR compared to those without, but the difference diminished over the course of therapy. Before anti-HCV therapy, patients with SVR had higher resistin levels than those without SVR ( diminished: the resistin levels of those with SVR decreased over the course of treatment. In contrast, the resistin levels in those without SVR fluctuated only slightly during the course of treatment, regardless of RETN score stratification ( Fig. 1b-d). Additionally, no differences in resistin level were noted between patients with or without IR, regardless of anti-HCV therapy (before anti-HCV therapy, p = 0.207; after anti-HCV therapy, p = 0.214). A pre-therapy resistin level > 3.956 ng/ml predicted SVR among patients with a RETN score of 0. To predict SVR, the RETN score was used to stratify a ROC analysis of pre-therapy resistin levels because it was associated positively with the resistin level (Tables 2 and 3) but negatively with the SVR rate in the multivariate analysis ( Table 4). The area under the ROC curve (AUC) for patients with a RETN score of 0 was 0.745 (p = 0.001, sensitivity = 82.57%, specificity = 60.00%), with a cut-off level of 3.956 ng/ml for pre-therapy resistin in determining SVR (Fig. 1e). In contrast, for those patients with a RETN score > 0, ROC analysis failed to provide a cut-off level of pre-therapy resistin in predicting SVR (RETN score > 0, i.e., score 1-4, AUC: 0.534, p = 0.5404) (Fig. 1f).   343) compared to those without SVR. As illustrated by serial sections from a representative case of a CHC patient with SVR, most intrahepatic resistin-positive cells were non-parenchymal cells (Fig. 2a,b, white arrows), including lymphocytes (Fig. 2c), HSCs (Fig. 2d), KCs (Fig. 2e) and HPCs (Fig. 2f). Some hepatocytes also expressed resistin (Fig. 2b, black arrows).

All 4 analyzed RETN SNPs were in linkage disequilibrium (LD) with each other; RETN-rs3745367 was in LD with IFNL3-rs12979860. Correlation tests showed the genotypes of the 4 aforementioned SNPs
to be tightly correlated (all p-values for the correlation tests for any 2 of the 4 RETN SNP genotypes were < 0.001). Furthermore, genotype "AA" of RETN-rs3745367 was associated with genotype "CC" of IFNL3-rs12979860 (linear-by-linear association p = 0.013, D' = 0.747) (Fig. 3a). These results indicate that all 4 RETN SNPs were in LD with each other and that RETN-rs3745367 was in LD with IFNL3-rs12979860.
Resistin reinforced IFNL3 expression. The fold changes in IFNL3 (Fig. 3b) or resistin (Fig. 3c) mRNA expression in PBMCs stimulated with recombinant proteins compared to PBMCs from the same individuals without stimulation are shown with their associated p-values in Fig. 3

Discussion
To our knowledge, this prospective study is the first to demonstrate the favorable role and associated basis for resistin in the response to interferon-based anti-HCV therapy. The five most compelling results are described below. (1) Among the 10 RETN SNPs investigated, rs34861192, rs3219175, rs3745367, and rs1423096 were synergistically associated with resistin levels in CHC patients. (2) In addition to these 4 RETN SNPs, pre-therapy ferritin, WBC count and GFR level were associated with pre-therapy resistin levels, whereas SVR, therapeutic  intervention, ALT and APRI levels, and WBC count were associated with the longitudinal trend in resistin levels.
(3) After adjusting for other confounders, resistin levels were positively associated with SVR, whereas the number of hyper-resistinemic genotypes of the 4 RETN SNPs (ie. RETN score) was negatively associated. SVR patients had significantly higher pre-therapy resistin levels than non-SVR patients, and a pre-therapy resistin level > 3.956 ng/ml predicted SVR in CHC patients with a RETN score of 0. (4) The differences in serum resistin level between SVR and non-SVR CHC patients mainly originated from intrahepatic lymphocytes, HSCs, KCs, HPCs, and hepatocytes. (5) Resistin may augment interferon-based anti-HCV therapy by reinforcing IFNL3. Genetically, RETN-rs3745367 was in LD with IFNL3-rs12979860. Functionally, both resistin and IFNL3 were upregulated, but only resistin was upregulated by recombinant resistin in PBMCs with RETN scores > 0 and = 0, respectively. HOMA-IR was not cross-sectionally or longitudinally associated with the resistin level in CHC patients, confirming that human resistin levels are not associated with glucose homeostasis but are associated with genetics (RETN SNPs) 14,15 , hepatic fibrosis/necroinflammation (ferritin) 22 , immunity (WBC counts) 23 , and renal function (GFR) 24 , even with HCV infection. Of note, only 4 of the 10 RETN SNPs documented in the literature 14,15 were synergistically associated with resistin levels in CHC patients. Of great importance, hyper-resistinemia is favorable for SVR. Following anti-HCV therapy, only SVR patients exhibited post-therapy increases in TC and TG levels along with decreases in ALT, APRI, and resistin levels. These findings reflect an improvement in HCV-associated hypolipidemia 2-4 , hepatic necroinflammation and fibrosis subsequent to viral clearance. However, an improvement in hepatic necroinflammation cannot fully explain the decreased resistin levels, as the difference in resistin levels between the SVR and non-SVR patients vanished at 24 weeks post-therapy, a time when the non-SVR patients still suffered from virus-associated necroinflammation. These results suggest that resistin may not only be pro-inflammatory but also immunomodulatory, augmenting interferon-based therapy to clear HCV. The relatively low pre-therapy resistin levels in the non-SVR patients indicate the inadequate responsiveness of immunity during interferon-based therapy. For CHC patients without hyper-resistinemic genotypes for the 4 RETN SNPs, in addition to the IFNL3 "CC" genotype, pre-therapy resistin levels > 3.956 ng/ml may serve as a feasible criterion for predicting favorable interferon-based therapeutic responses. Indeed, hyper-resistinemia in patients with CHC is consistently observed 25,26 , is reversed after viral clearance 27 , and causes moderate to-severe fibrosis 25 but is thought not to be associated with therapeutic response 27 . The beneficial role of hyper-resistinemia in the response to interferon-based therapy could not be revealed without eliminating renal insufficiency, which profoundly increases resistin levels 24 , and without adjusting for RETN SNPs in over 500 CHC patients undergoing anti-HCV therapy.
The CHC patients exhibited higher expression of both PB and hepatic resistin than the controls. However, the difference in serum resistin levels between the SVR and non-SVR CHC patients primarily originated from the liver. This result highlights the importance of intrahepatic resistin in augmenting interferon-based anti-HCV efficacy and is consistent with the hepatotropic characteristics of HCV. All of the intrahepatic resistin-positive cells, including lymphocytes, HSCs, KCs, HPCs, and hepatocytes, are reported to express resistin [28][29][30][31] . Resistin may become available to resident hepatocytes through paracrine mechanisms via local production by lymphocytes, HSCs and KCs 30,32 , thereby accelerating hepatic necroinflammation and fibrosis 33 . Consistently, resistin  Spearman's rank correlation coefficients are presented using a spectrum of colors ranging from dark brown to deep purple (the color bar is shown at the right). The HCV genotypes include genotype type 1 (coded as 1) and non-genotype type 1 (coded as 2). SVR: sustained viral response. SNPs are ordered by their locations along chromosome 19. The genotypes of 4 RETN SNPs are highly correlated with each other, forming a mosaic red-brown color block near the center of the color map. IFNL3 genotype "CC" is strongly correlated with RETN-rs3745367 genotype "AA" (Spearman correlation coefficient = 0.172, p = 0.013). (b,c) RNA fold changes for IFNL3 (b) and resistin (c) after stimulation with recombinant resistin (green bars) or IFNL3 (red bars) in peripheral blood mononuclear cells (PBMCs). levels correlate with inflammatory markers in chronic liver disease 34 and unfavorable outcomes in non-septic critical patients 35 . In the current study, longitudinal resistin levels correlated with APRI, and both HSCs and HPCs expressed resistin. HPC activation occurs in response to hepatic oxidative stress and is associated with fibrosis 31 , as evidenced by HSC involvement. Altogether, the pro-inflammatory activities of resistin are harmful if left untreated, particularly with respect to accelerating hepatic fibrosis. In addition, viral clearance following interferon-based therapy is associated with the induction of a vigorous T-helper response 36 . KCs, HSC, and hepatocytes may serve as liver antigen-presenting cells (APCs) to induce primary T-cell activation 37 . Resistin is upregulated during monocyte-macrophage differentiation 7 and induces NF-kB activity in mature T-helper cells, thereby enhancing cell chemotaxis 38 . Thus, the fact that KCs, lymphocytes, HSCs, and hepatocytes all expressed resistin suggests that resistin might initiate innate immunity with macrophage activation to invoke adaptive immunity, whereby a reinforcing T-helper response boosts the efficacy of interferon-based therapy. In addition, IFNL3 is expressed predominantly in APCs and is upregulated by interferons 39 , and IFNL3 SNP genotypes are highly associated with the efficacy of interferon-based therapy in CHC 40 . However, IFNL3 SNPs do not affect expression of the IFNL3 gene itself, and the link between an IFNL3 SNP-associated therapeutic response and IFNL3 expression remains elusive 41 . The well-established LD between RETN and IFNL3 SNPs in the present study indicates that the aforementioned elusive link may be related to complicated genetic accommodation 42 . Namely, two correlated genes, RETN and IFNL3, might together modulate host immunity against HCV. Moreover, functional connections between IFNL3 and resistin were identified. Both resistin and IFNL3 mRNAs were upregulated, but only resistin mRNA was upregulated by recombinant resistin in PBMCs with and without hyper-resistinemic genotypes of the 4 RETN SNPs, respectively. In effective IFNL3 reinforcement, more benefits appeared to be obtained by the CHC patients with RETN hyper-resistinemic genotypes than without. However, this benefit might be compromised or even reversed in viral clearance, as the presence of hyper-resistinemic genotypes suggest aggravated hepatic necroinflammation/fibrosis subsequent to high baseline resistin levels. The deteriorated milieu of hepatic necroinflammation, such as severe hepatic fibrosis, is actually detrimental for HCV clearance 43 . This likely accounts for why pre-therapy resistin levels failed to predict the response rate among those with RETN hyper-resistinemic genotypes, and RETN score was negatively associated with SVR. The proposed basis of the connection between IFNL3 and resistin and associated anti-viral pathways is shown in Fig. 3d.
In an era in which most HCV infections can be eradicated with potent, direct-acting anti-viral agents 44 , the significance of RETN SNPs and resistin is still important in low-resource settings, where interferon-based therapy remains the mainstay treatment for CHC patients. Moreover, such findings may be extended for evaluating or improving therapeutic efficacy for other viral infections such as hepatitis B virus, for which interferon-α and nucleoside/nucleotide analogs are currently available but unsatisfactory 45 . In particular, interferon-α is effective in only approximately 30% of chronic hepatitis B cases 46 . Regardless, any anti-viral strategy targeting resistin-associated pathways demands extreme caution due to the double-edged nature of resistin, namely, necroinflammation and immunomodulation.
Taken together, the 4 RETN SNPs, rs34861192, rs3219175, rs3745367 and rs1423096, were synergistically associated with resistin levels. Fine-tuned by the 4 RETN SNPs, pre-therapy resistin levels positively affected the response to interferon-based anti-HCV therapy. The liver is the main source of resistin in the multi-cellular orchestration of viral clearance. Reinforcement of IFNL3 by resistin was evident at both the genetic and functional levels. This discovery of the beneficial role of resistin in the interferon-based anti-HCV therapeutic response encourages the development of novel anti-viral agents targeting resistin-associated pathways.