Sofosbuvir Is the Major Contributor to Elevation of Blood Cholesterol and Low-density Lipoprotein During HCV Treatment With DAAs

Background and aims: Worsened lipid proles were observed in chronic hepatitis C (CHC) patients during direct-acting antivirals (DAAs) treatment, among which combination drugs confounded the effect of individual ingredient on lipid. Tenofovir alafenamide (TAF) also worsened lipid proles in HIV patients. Structural similarity between sofosbuvir (SOF) and TAF prompted us to investigate rapid increase in total cholesterol (TC) and low-density lipoprotein (LDL) in CHC patients treated with SOF-based DAAs. Methods: A retrospective study was performed to analyze 487 CHC patients receiving DAAs with SVR12. Laboratory data were analyzed by logistic regression to determine relative risks of TC and LDL of SOF-based regimens over non-SOF-based regimens. Week 4 to baseline ratios of serum TC or LDL between regimens were compared by Mann-Whitney's test. Results: 487 patients were treated with Harvoni ® (SOF-based, 206 patients), Epclusa ® (SOF-based, 124 patients), Maviret ® (non-SOF-based, 122 patients), or Zepatier ® (non-SOF-based, 35 patients). At week 4 during drug treatment, Harvoni ® , Epclusa ® , and Maviret ® induced statistically signicant elevation of TC and LDL, but Zepatier ® did not. SOF-based regimens had 2.72-fold higher relative risk (RR) causing 10% elevation of TC (95% CI: 1.84-4.02, p < 0.001) and 2.04-fold higher RR causing 10% elevation of LDL (95% CI:1.39-3.01, p < 0.001) than non-SOF-based DAAs. Conclusion: SOF-based DAAs were associated with signicantly larger amplitude of increases in TC and LDL than non-SOFO-based regimens during the initial 4 weeks of treatment, but the amplitude was not sustained to SVR12.


Introduction
Hepatitis C virus (HCV) is one of the major causes of liver-related morbidity and mortality,(1) estimating infection of approximately 180 million people worldwide. (2) The HCV life cycle is initiated by binding of virus particles to hepatocellular receptors, endocytosis, fusion of HCV glycoproteins with endosomal membranes, acidi cation of endosome, and release of the viral genome into cytosol for replication (3).
Internal ribosome entry site (IRES)-mediated translation of incoming viral RNA enables viral gene expression and processing, and replication occurs in the "membranous webs" (4). Following replication, genomic RNAs in complex with NS5A protein transit to lipid droplets, where core protein localizes and virion assembly occurs (5). After acquiring apolipoproteins B and E (apoB and apoE), components of VLDL (very low-density lipoproteins) and LDL, HCV infectious particles egress in a manner that parallels the VLDL secretory pathway(6, 7). HCV-infected patients have a prevalence of hepatic steatosis twofold higher than in HBV-infected patients,(8, 9) demonstrating a clear correlation between HCV infection and non-alcoholic fatty liver disease. These patients are also more likely to present decreased serum levels of apoB-bearing lipoproteins because HCV seizes these lipoproteins. (10,11) HCV nonstructural proteins 4B, 5A and 5B constitute a complex for RNA replication. The RNA translates into a polyprotein processed by viral NS3/4A and host proteases to generate structural proteins for viral assembly and nonstructural proteins involved in RNA replication. (12) Currently, direct-acting antivirals (DAAs), which have replaced interferon, are the standard treatment for HCV infection. (13,14) According to mechanisms of action and therapeutic targets, DAAs are classi ed into four categories: nonstructural proteins 3/4A (NS3/4A) protease inhibitor, NS5A replication complex inhibitor, and NS5B nucleoside and non-nucleoside polymerase inhibitor. (15) In the class of NS5B nucleoside polymerase inhibitor, sofosbuvir (SOF) is the only drug and plays an important role in the combination of other DAAs for HCV treatment. (13,14) Although DAAs provide well-tolerated, safe, and highly e cacious outcomes,(16) several studies reported HCV virion is tightly associated with hepatocyte-derived lipoproteins to form a lipid-laden particle, called lipo-viro-particle. It was thought that HCV hijacks lipoproteins that are released to the blood after clearance of the virus by DAAs. Given that the concept is true, all DAA regimens would have equivalent effects on lipid pro les of the blood. In this report, we performed retrospective analysis based on structural similarity between TAF (tenofovir alafenamide) and SOF (Fig. 1A) and the observation that patients given TAF for HIV (human immunode ciency virus) infection had elevated cholesterol level (19)(20)(21). Therefore, we hypothesized that SOF may be the major contributor to cholesterol and LDL elevations during DAAs treatment for patients infected by HCV. The mechanism may be related to the phosphoramidate side chain that is cleaved from sofosbuvir to produce an active nucleotide analogue, GS-461203 (2'-deoxy-2'-α-uoro-β-C-methyluridine-5'-triphosphate). (22) (Fig. 1B)

Study population
The study retrospectively enrolled HCV-infected patients who received DAAs treatment for 8 or 12 weeks at Taipei Veterans General Hospital from September 2018 to August 2020. All patients were aged above 20, male or female and had chronic HCV infection, de ned as detectable anti-HCV antibody and HCV RNA level in the serum for more than 6 months. Patients were excluded from the study if they had decompensated cirrhosis. The study was approved by the Institutional Review Board of Taipei Veterans General Hospital and was conducted in accordance with the principles of Declaration of Helsinki and the International Conference on Harmonization for Good Clinical Practice. All patients read and signed informed consent before drug prescription and study-related procedure.

Study design
Baseline demographic data were collected before treatment. Hemogram, serum biochemical pro les (albumin, total bilirubin, direct bilirubin, aspartate aminotransferase [AST], alanine aminotransferase

Virologic assessment
On-treatment effectiveness was assessed by detecting serum HCV RNA levels at weeks 4 and 12. The effectiveness at the end of treatment was SVR12, de ned as serum HCV RNA level < LLOQ 12 weeks after completed treatment.

Statistical analyses.
All analyses were performed using STATA (12th ed., developed by StataCorp LLC, College Station, TX, USA). The pre-treatment patient characteristics were shown in median (range) and percentages as appropriate and compared by Mann-Whitney's test and χ2 with Fisher's exact test. The effectiveness of treatment during and after drug administration was shown in number and percentages. Week 4 to baseline ratios of serum cholesterol or LDL between regimens were compared by Mann-Whitney's test or logistic regression. All statistics were two-tailed, and the results were considered statistically signi cant when a p value was < 0.05.

Patient characteristics
Four hundred and eighty-seven CHC patients achieving SVR12 were included in the study ( Table 1). The median age was 63 years, and 247 patients were male (50.7%). Four hundred and fty-one patients were treatment-naïve (92.6%), and 35 had HBV co-infection (7.2%). No patient received antiviral therapy for HBV, and neither developed HBV reactivation nor hepatitis ares during DAA therapy. Forty-one patients had HIV co-infection, who took anti-HIV drugs. The median log 10 HCV RNA level was 6.18. One hundred and ninety-six patients had a baseline viral load < 800,000 IU/mL (40.3%). One hundred and one patients had a Fib-4 brosis stage of F3 (20.8%), and 54 patients, F4 (11.1%) ( Table 1).

Amplitudes of changes in TC and LDL between SOF-and non-SOF-based regimens
Amplitudes of changes in serum TC and LDL from baseline to week 4 (or SVR12) were expressed as Log 10 [(Week 4 or SVR12)/Baseline] and shown at Y axis of Fig. 3. SOF-based regimens caused signi cantly larger amplitude of change in TC (Fig. 3A, p < 0.001) and LDL (Fig. 3B, p < 0.001) at week 4 compared to non-SOF-based regimens. These changes were not sustained at SVR12 (TC, p = 0.884; LDL, p = 0.475). The strati ed analyses compared different two regimens and showed that there was no signi cant difference in log 10 [Week4/Baseline] change of both TC (Fig. 3C, p = 0.361) and LDL (Fig. 3D, p = 0.248) between the two SOF-based regimens, SOF/VEL and SOF/LED. The statistically signi cant changes in TC and LDL were noted at these two-regimen comparisons: SOF/VEL vs GLE/PIB (TC, p = 0.001; LDL, p = 0.010), SOF/LED vs GLE/PIB (TC, p < 0.001; LDL, p < 0.001), and SOF/LED vs ELB/GRA (TC, p < 0.001; LDL, p = 0.011). The p value of TC at SOF/VEL vs ELB/GRA was 0.017, and that of LDL was 0.070. These data suggested that SOF was associated with the major inductor of higher week-4 lipid change.
Incidence comparisons of increased TC or LDL in SOF-and non-SOF-based regimens.
There were no signi cant differences in week-4 TC and LDL elevations between two SOF regimens (SOF/VEL and SOF/LED) and between two non-SOF regimens (GLE/PIB and ELB/GRA) in either increase > 10% or >25%. However, in TC > 10%, all SOF vs non-SOF (SOF/VEL vs GLE/PIB, SOF/VEL vs ELB/GRA, SOF/LED vs GLE/PIB, and SOF/LED vs ELB/GRA) showed signi cant differences in TC elevation. In LDL > 10%, SOF/LED caused signi cant higher risk in LDL elevation than GLE/PIB and ELB/GRA. In TC or LDL > 25%, SOF-based regimen still showed signi cant higher risk in some SOF vs non-SOF paired comparisons. These statistics gave further evidence for SOF in worsened week-4 lipid pro les. The SOFbased lipid-elevation effects were not seen at SVR12 (Supplementary Table 1).

Discussion
DAAs have been shown to worsen lipid pro les during CHC treatment. To the best of our knowledge, this study is the rst retrospective analysis to evaluate the effect of SOF-based DAAs on changes in lipid pro les. Our results showed that compared with non-SOF DAAs, SOF-based DAAs were associated with more rapid increases in TC and LDL during the initial 4 weeks of treatment than the non-SOF counterparts. However, the SOF-potentiated effects were not sustained at the end of treatment. A further comparison of SOF/VEL or SOF/LED with GLE/PIB and ELB/GRA revealed a similar trend. It is interesting to note that ELB/GRA did not cause TC and LDL elevation at week 4. Very limited literature documented association of ELB/GRA with lipid worsening during CHC treatment. Sun et al. reported 24 cases, 13 treated with ELB/GRA and 11 with SOF/LED, showing signi cant elevation of cholesterol at week 4(23), which we think to be the effect of SOF/LED rather than ELB/GRA. However, at SVR12, TC and LDL tended to rise in ELB/GRA group compared with its baseline (TC, p = 0.035; LDL, p = 0.037, Fig. 2).
Combination DAAs have been the treatment standard for CHC. Our data provide strong evidence that SOF-based DAAs resulted in higher elevation of TC and LDL than non-SOF-based regimens at week 4 after drug administration. Since the two kinds of regimens achieved similar SVR12, viral clearance appeared not to be the sole mechanism to account for different TC and LDL changes between the two. We compared three compounds, sofosbuvir, tenofovir alafenamide (TAF), and tenofovir disoproxil fumarate (TDF) and found that SOF and TAF have similar phosphoramidate side chains (Fig. 1A). Milinkovic et al. reported changes in lipid pro le in HIV patients treated with TDF or TAF. After switching from TDF to TAF, mean total cholesterol increased from 186 ± 37 mg/dl at baseline to 206 ± 43 mg/dL and 204 ± 43 mg/dL at weeks 12 and 24 (p < 0.001), and the increase in total cholesterol was mainly due to an increase in LDL-cholesterol. (20) Similar lipid changes were reported by other HIV investigators. (21,24) SOF and TAF are nucleotide prodrugs that are converted to active compounds by enzymatic cleavage of phosphoramidate side chain. SOF decomposed to GS-060965, phenolate ion, and propan-2-yl 2aminopropanoate within hepatocytes (Fig. 1B). (25) Since TDF does not cause change in lipid as do SOF and TAF, the cleaved products from phosphoramidate may be the agents that promote b-lipoprotein synthesis and secretion by hepatocytes.
For CHC patients, SOF-based DAAs rapidly and signi cantly elevated TC and LDL levels during the initial 4 weeks of treatment, but the higher amplitudes of changes in TC and LDL tended to disappear at the end of treatment. It could be deduced from the chemical structures of SOF and TAF that the cleaved products phenolate ion and propan-2-yl 2-aminopropanoate from the phosphoramidate side chain may promote blipoprotein synthesis and secretion initially and then induce more enzymes to destroy them with time, so TC and LDL levels jump transiently and then decline.
Our results showed trends of increases in TC and LDL at week 4 and SVR12. Some reports showed the increases in TC or LDL disappeared after treatment. (26,27) Other studies suggested the elevated LDL continued to post-treatment one year.(28-31) The main reason for the inconsistency was that most of these studies were single-arm studies without placebo controls. Additionally, both Younossi et al. and Pedersen et al. observed that genotype 3 patients had signi cantly increased LDL during DAAs treatment, but genotype 1 or genotype 2 patients did not. (32,33) As the majority of the included patients in the present study were genotype 1 (59.6%) and genotype 2 (32.9%) patients (Table 1), changes in LDL would be less, consequently reducing the difference between treatment groups. Furthermore, genetic factors have been reported in association with changes in LDL. In the study by Emmanuel et al., the IFNL4-ΔG carriers had signi cant increases in LDL during DAAs treatment and at post-treatment one year, but the patients with IFNL4-TT/TT did not.(34) In the study by Morihana et al., the difference in LDL between sofobusvir/ledipasvir and daclatasvir/asunaprevir disappeared after the end of treatment. However, the IL28B TG/GG patients continued to have increased LDL from the end of treatment to post-treatment two years, whereas the IL28B TT patients did not.(26) Because the current analysis did not consider these genetic factors, our results might be potentially confounded by these predictors.
Although it was common to observe increases in TC and LDL during DAAs treatment, a few studies analyzing risk factors for increased LDL during DAAs treatment revealed discordant ndings. In the study by Morihana et al., multivariate regression analysis showed that changes in LDL were negatively correlated with baseline LDL and HDL.(26) Hashimoto et al. reported that the decline of HCV core antigen from day 0 to day 1 was independently associated with amplitudes of changes in LDL but not with baseline LDL during DAA treatment.(35) The discordance highlighted the need of prospective large-scale long-term studies to elucidate the effect of DAAs on lipid pro les.
There were several limitations in this study. Firstly, lipid pro les are incomplete without HDL and triglyceride. Secondly, case numbers are uneven among DAA regimens. The third is that the included patients in this study were mainly those who achieved SVR12. In addition, we did not have lipid data of post-treatment long-term follow-up. Therefore, it deserves more comparative studies, which evaluate complete lipid pro les at different time points among DAAs regimens.
Declarations ACKNOWLEDGEMENTS This study was supported by grants from the Ministry of Science and Technology, Taiwan (MOST-106-2314-B-075-055 to Keng-Hsin Lan; MOST-107-2314-B-075-049 -MY2 to Wei-Ping Lee), and Taipei Veterans General Hospital (V107C-137 to Keng-Hsin Lan; V108C-206 to Wei-Ping Lee), Taipei, Taiwan. . Clinical Trials Registration: This is not a clinical trial study 7. Author Contribution: All authors have contributed substantially to preparation of this manuscript including conception and design, execution, analysis, and interpretation of data, and drafting the article.
. Con ict of Interest: All authors listed declare no con ict of interest. Structures of tenofovir alafenamide (TAF), sofosbuvir (SOF), and tenofovir disoproxil fumarate (panel A).