CD2-Associated Protein Contributes to Hepatitis C, Virus Propagation and Steatosis by Disrupting Insulin Signaling

Materials

Cell cultures, viruses, plasmids, and reagents are described in the Supporting Information.

Animal Model

Eight- to 12-week-old age- and sex-matched littermates of human CD81 and occludin transgenic ICR mice were used for in vivo study and infected with HCV as described.15 Mice were tail-vein injected with HCV J399EM (tissue culture infective dose, 50 = 1 × 10⁸/mL; 1 mL in 1-2 minutes to avoid liver injury). Mouse blood (0.1 mL) and liver tissues (0.1 g) were collected to quantify HCV genomic RNA at the indicated time. Five mice at each time point were infected with HCV. Two to three HCV infection-confirmed mice were used for further analysis. One of the noninfected mice at each time point was used as negative control. Data collection and data analysis were performed by different persons in a blinded manner. Use of animals was approved by the Institutional Review Board of Wuhan Institute of Virology, Chinese Academy of Sciences (Wuhan, China).15

Human Subjects

Seventy-two serologically confirmed HCV-infected human liver biopsies were obtained from resected liver tissues containing HCC, hemangioma, or cholangiocarcinoma from patients at the Tongji Hospital (Wuhan, China) and Eastern Hepatobiliary Surgery Hospital (Shanghai, China; patients’ information in Supporting Table S1). No biopsies were obtained from executed prisoners or other institutionalized persons. Liver samples from HCV/HBV (hepatitis B virus) coinfection were excluded. Twelve non-HCV- and non-HBV-infected control specimens were obtained from normal regions of liver adjacent to resected hemangioma (patients’ information in Supporting Table S1). Biopsies were scored for steatosis, cirrhosis, and HCC by two pathologists, Changshu Ke (M.D., Ph.D.) and Yu Hu (M.D., Ph.D.; Department of Pathology, Tongji Hospital). Among the 72 biopsies, 53 also had HCC, 7 cases did not show steatosis and cirrhosis, 17 cases showed only steatosis, 20 cases showed steatosis and cirrhosis, and 4 cases showed only cirrhosis. Some cases could not be determined and thus were excluded. Informed consent was obtained from all subjects. Use of liver sections was approved by the Institutional Review Board of Wuhan Institute of Virology, Chinese Academy of Sciences (Approval Number: WIVH28201601).

Methods of IN VITRO Assays, Statistical Analysis

In vitro assays and statistical analysis are described in the Supporting Information.

HCV NS5A BINDS CD2AP

HCV NS5A plays an important role in HCV propagation. To identify proteins participating in HCV propagation, Huh7 cells with the NS5A-BirA*-HA construct (Fig. 1A) were cultured with or without biotin; more biotin-labeled proteins were detected in cells cultured with biotin (Fig. 1A). Several bands presented only in samples with biotin were sequenced and identified (Supporting Table S2). Among those proteins, CD2AP is an adaptor protein with three Src homology 3 (SH3) domains. Conversely, NS5A has several proline-rich motifs that are reported to bind the SH3 domain.16 Structural information thus suggests that CD2AP might interact with NS5A. Furthermore, the association of HCV replication complexes with actin has been shown depending on NS3 and NS5A interaction,17 and CD2AP can regulate actin organization.18 We thus focused on studying interactions between CD2AP and NS5A.

We found that CD2AP indeed interacted with NS5A (Supporting Fig. S1A), but not core protein (Supporting Fig. S1B) in vitro. In HCV Japanese fulminant hepatitis type 1 (JFH1)-infected cells, anti-CD2AP antibody pulled down NS5A and CD2AP (Fig. 1B). Conversely, anti-NS5A antibody also pulled down CD2AP and NS5A (Fig. 1B). Copurification was not observed with an isotype control mouse immunoglobulin (Ig)G2a (Fig. 1C). In addition, we observed colocalization of CD2AP and NS5A in cells infected with HCV-J399EM (Fig. 1D), the HCV 2a strain in which NS5A is enhanced green fluorescent protein tagged. Thus, in HCV-infected cells, CD2AP interacts with NS5A. More important, levels of CD2AP were consistently up-regulated in infected cells compared to mock-infected cells (Fig. 1B).

Domain III of NS5A Interacts with SH3 Domain of CD2AP

To investigate which domain bound NS5A, a panel of FLAG-tagged CD2AP mutants was generated (Fig. 1E) and coexpressed with hemagglutinin (HA)-tagged NS5A in 293T cells. By co-immunoprecipitation (Co-IP), CD2AP containing at least one SH3 domain was able to bind NS5A, but CD2AP mutant lacking all three SH3 domains did not (Fig. 1E). NS5A contains an N-terminal amphipathic helix, three domains (I, II, and III), separated by two low-complexity sequences.19 We generated a series of NS5A mutants and found that when domain III of NS5A was deleted, it failed to bind CD2AP (Fig. 1F). Therefore, domain III of NS5A can bind the SH3 domains of CD2AP.

CD2AP Participates in the Transport of NS5A to LD

To investigate the function of the binding between CD2AP and NS5A, we expressed an mCherry-tagged full-length CD2AP or CD2AP lacking all SH3 domains in Huh7.5.1 cells. After infecting these cells with HCV-J399EM, we observed that full-length CD2AP, but not truncated CD2AP, colocalized with NS5A (Fig. 2A). We then performed live cell tracking of the colocalized NS5A/CD2AP complex. The trajectory and velocity of the complex supported that CD2AP and NS5A indeed comigrated (Fig. 2B; Supporting Movie S1).

HCV particles travel along actin filaments.20 Motility of CD2AP in podocytes depends on actin polymerization.21 Others have reported that transport of the HCV NS5A to LDs depend on microtubules.12, 17, 22 To address whether cotransport of NS5A and CD2AP was actin or microtubule dependent, we cultured HCV-infected cells with either cytochalacin B (an actin polymerization inhibitor) or colchicine (a microtubulin polymerization inhibitor). Live cell tracking of NS5A and CD2AP showed that cytochalacin B reduced the colocalization of NS5A and CD2AP (Fig. 2C). However, colchicine did not reduce the colocalization, but blocked their co-movement (Fig. 2D; Supporting Movie S2). Four hours after removing cytochalacin B, colocalization of CD2AP and NS5A was restored (Fig. 2C). However, removal of colchicine did not restore NS5A/CD2AP comigration (movie not shown). These results suggest that NS5A and CD2AP colocalization is actin dependent and cytochalacin B treatment is reversible. Co-movement of the NS5A/CD2AP complex is microtubule dependent and colchicine treatment is irreversible.

Next, we investigated whether CD2AP participated in transport of NS5A to LD. We established two CD2AP-silenced Con1 cell lines (Supporting Fig. S2, c4# and c6#). Con1 cells support the subgenomic replication of HCV RNA, but do not support HCV assembly and production of virus particles.23 Silencing CD2AP did not affect the total level of NS5A, but reduced the association of NS5A with fractionated LDs (Fig. 2E). Purity of LDs was confirmed by the absence of calnexin, a marker for endoplasmic reticulum (ER), and the presence of adipose differentiation-related protein, a marker for LDs (Fig. 2E). Thus, these results suggest that binding between NS5A and CD2AP is actin polymerization dependent and NS5A/CD2AP complex may be transported to LDs in a microtubule-dependent manner.

Silencing CD2AP Reduces the Level of LDs

Because down-regulation of CD2AP diminished NS5A association with LDs in Con1 cells, we next investigated whether the phenomenon could be observed in Huh7.5.1 cells infected with HCV by generating two stably CD2AP-silenced cell lines (Supporting Fig. S3A, 4# and 6#). Immunofluorescent staining showed that less NS5A was found on LDs in CD2AP-silenced cells after infection with HCV (Fig. 3A). Surprisingly, levels of LDs were also reduced after CD2AP silencing (Fig. 3A). To investigate whether LD levels are enhanced by CD2AP expression or simply attributed to HCV infection, we stained LDs in those cells without HCV infection. We found that the level of LDs was reduced in CD2AP-silenced cells treated with or without oleic acid (OA), which can stimulate LD formation by activating the long-chain fatty acid receptor, free fatty acid receptor 4 (Fig. 3B).24 To prove there is a functional interplay between CD2AP and LD accumulation, we reintroduced HA-tagged CD2AP back into CD2AP-silenced cells (Supporting Fig. S3B). CD2AP-rescued cells (6#-res) had more LDs than control (6#-NC; Fig. 3C). Therefore, in addition to targeting HCV components to LDs, CD2AP also plays a role in boosting LD level.

CD2AP Increases Insulin Receptor Substrate 1 Ubiquitination by Engaging Casitas B-Lineage Lymphoma (b) E3 Ubiquitin Ligases

We then investigated the mechanism of CD2AP regulating LD accumulation. Level of LDs is modulated by insulin signaling.25, 26 To study whether CD2AP boosted LD level through insulin signaling, we detected the expression of insulin receptor and insulin receptor substrate 1 (IRS1) in CD2AP-silenced cells. We found that levels of total IRS1, but not total insulin receptor, were obviously up-regulated (Fig. 4A). However, there was no difference for IRS1 mRNA levels between control and CD2AP-silenced cells (Supporting Fig. S4A). Hence, expression of IRS1 is likely regulated at the protein level. We found that after 2 hours of treatment with proteasome degradation inhibitor MG132, but not other inhibitors, such as NH₄Cl/3-methyladenine/chloroquine, level of IRS1 was increased (Supporting Fig. S4B). These results suggest that IRS1 stability may be regulated by protein ubiquitination. Furthermore, MG132 treatment increased much more IRS1 in wild-type cells than in CD2AP-scilenced cells (Fig. 4B). These results implicate that CD2AP down-regulation may diminish IRS1 ubiquitination. To prove this hypothesis, we purified a similar amount of IRS1 from CD2AP-silenced and control cells, and found that CD2AP down-regulated cells indeed showed less ubiquitination (Fig. 4C). Thus, CD2AP affects IRS1 level by modulating its ubiquitination. However, we could not exclude the possibility that some IRS1 might be degraded in a ubiquitin-independent manner.

We then investigated how CD2AP regulated IRS1 ubiquitination. In Huh7.5.1 cells, IRS1 copurified and colocalized with CD2AP (Fig. 4D). Because IRS1 is a client of several E3 ubiquitin ligases27, 28 and CD2AP is known to interact with Cbl family E3 ligases,29, 30 we then investigated whether IRS1 interacted with casitas B-lineage lymphoma (b) (Cbl/Cbl-b) E3 ligases. We found that Cbl/Cbl-b copurified and colocalized with IRS1 (Fig. 4D). To address whether Cbl/Cbl-b ubiquitinated IRS1, we silenced Cbl/Cbl-b in Huh7.5.1 cells. Accordingly, levels of IRS1 were up-regulated (Fig. 4E). Thus, CD2AP promotes IRS1 ubiquitination and degradation by engaging Cbl/Cbl-b E3 ligases.

CD2AP Disrupts IRS1 Signaling to Boost LD Level

Because CD2AP promotes IRS1 degradation, we then investigated whether CD2AP influenced LD formation by the IRS1 signaling pathway, which affects lipid metabolism through the IRS1/phosphatidylinositol 3-kinase (PI3K)/protein kinase B (Akt) axis.31, 32 In CD2AP-silenced cells, the level of phosphorylated (p)-Akt (S473), but not p-Erk (extracellular signal-regulated kinase) was increased (Fig. 5A). Akt is a negative regulator of adenosine monophosphate kinase (AMPK).33 Then, we blotted for AMPK and found that level of p-AMPK (T172) was reduced in CD2AP-silenced cells (Fig. 5A). Activation of AMPK regulates many enzymes in lipid metabolism, such as acetyl-CoA carboxylases 1 and 2 (ACC1/2), adipose triglyceride lipase (ATGL), sterol regulatory element-binding protein 1 (SREBP-1), and hormone-sensitive lipase (HSL).34, 35 Binding of HSL to p-perilipin A on the surface of LD causes hydrolysis of LDs.36 AMPK phosphorylates S554, inactivating the lipase activity of HSL.37 We found that levels of p-HSL (S554), a marker of HSL inactivation, were reduced in CD2AP-silenced cells (Fig. 5A). Furthermore, silencing CD2AP did not affect other proteins involved in lipid metabolism, such as p-ACC1/2, ACC1/2, SREBP1, and ATGL (Supporting Fig. S5). To confirm that HSL (S554) is phosphorylated by AMPK, we treated cells with dorsomorphin, an AMPK inhibitor. Dorsomorphin reduced levels of p-AMPK (T172) and p-HSL (S554; Fig. 5B). To further prove that CD2AP disrupts IRS1 signaling by the PI3K-Akt pathway, we reintroduced CD2AP into CD2AP-silenced cells; the level of p-Akt (S473) was reduced (Fig. 5C, compare lane 3 to lane 4). Considering that CD2AP functionally rescued p-Akt (Fig. 5C) and LD levels (Fig. 3C) when reintroduced into CD2AP-silenced cells, these results provided strong evidences that enhanced CD2AP affects lipid metabolism by the IRS1-Akt-AMPK-HSL axis, causing inactivation of HSL to increase LD accumulation.

CD2AP Silencing Inhibits HCV Propagation Independent of Virus Entry, Subgenomic Replication, and Internal Ribosomal Entry Site-Dependent Translation

Because silencing CD2AP affects LD accumulation and NS5A transport to LDs, we posited that CD2AP down-regulation should affect HCV propagation. First, we infected CD2AP-silenced Huh7.5.1 cells with HCV-JFH1 and found that CD2AP down-regulation did not affect cell viability (Supporting Fig. S6A), but reduced intracellular total HCV-RNA level, protein level, and supernatant RNA copy numbers (Fig. 6A), and the reduction seems to be correlated with level of CD2AP reduction. In addition, silencing CD2AP also decreased the JFH1-Luc-5AGFP reporter virus activity (Fig. 6B) and CD2AP-rescued cells (6#-res) could recover HCV-RNA and protein level (Fig. 6C). Thus, CD2AP indeed regulates HCV propagation.

We then tried to identify the step(s) affecting HCV replication after CD2AP silencing. We found that silencing CD2AP did not affect HCV entry (by pseudovirus assay; Supporting Fig. S6B), subgenomic replication in Con1 cells, which do not support HCV assembly (Supporting Fig. S6C and Fig. 2E), and HCV internal ribosomal entry site-dependent translation (Supporting Fig. S6D). However, intracellular and supernatant HCV titers (Fig. 6D) were reduced in CD2AP-silenced cells infected with JFH1. This suggested that CD2AP silencing impaired production of infectious virions in steps other than virus entry, subgenomic replication, and RNA translation. We further performed one single-cycle assay to determine which step of HCV infection was affected by CD2AP silencing. We found that HCV could complete one single cycle between 12 and 18 hours postinfection (Supporting Fig. S7A) and CD2AP silencing significantly reduced intracellular HCV RNA 18 hours, but not 12 hours, postinfection (Fig. 6E). However, we did not detect NS5A protein expression until 24 hours postinfection (Supporting Fig. S7B), probably attributed to low sensitivity of immunoblotting. These results, together with the observation that less LD accumulation and less NS5A was found on LDs after CD2AP silencing (Fig. 3), supported that HCV assembly might be impaired in CD2AP-silenced cells. However, it is also possible that steps other than HCV assembly are also impacted by CD2AP down-regulation. Thus, HCV infection increased CD2AP expression to enhance HCV propagation.

CD2AP Up-Regulation Correlates with Steatosis in an Animal Model and in Human Biopsies

HCV is known to cause steatosis, and CD2AP is up-regulated in HCV-infected cells and enhanced CD2AP could stimulate LD accumulation, a requirement for steatosis. We then investigated whether steatosis and CD2AP up-regulation could be observed in a human CD81 and occludin transgenic ICR mouse model.15 Based on liver and blood HCV-RNA copy number, the first week after HCV infection represented the acute infection stage (Fig. 7A). At this time, there was little CD2AP staining in the liver (Fig. 7B). In contrast, CD2AP expression became highly noticeable from the first to fourth month during the chronic infection stage, when lipids began to accumulate and steatosis was evident (Fig. 7B, arrows indicate steatosis). Thus, up-regulation of CD2AP correlates with manifestation of steatosis.

Finally, we stained 72 liver biopsies from HCV-infected patients and 12 liver biopsies from non-HCV-infected controls for CD2AP. We found CD2AP immunohistochemistry (IHC) staining was faint or undetectable in most of the non-HCV-infected controls, whereas most of the 72 HCV-infected biopsies showed positive CD2AP staining (Fig. 7C,D). Most important, HCV-infected biopsies showing only steatosis or HCC also showed stronger CD2AP staining than non-HCV-infected hepatocytes (Fig. 7D). Thus, HCV infection activated CD2AP expression in the mouse model and human subjects, and the up-regulation correlates well with liver steatosis (LS).