Simultaneous detection of hepatitis C virus and interferon stimulated gene expression in infected human liver

Authors

  • Stefan Wieland,

    Corresponding author
    1. Department of Immunology and Microbial Science, Scripps Research Institute, La Jolla, CA
    • Address reprint requests to: Stefan Wieland or Francis V. Chisari, Department of Molecular and Experimental Medicine, Scripps Research Institute, 10550, North Torrey Pines Road, La Jolla, CA 92037. E-mail: swieland@scripps.edu or fchisari@scripps.edu; fax: 858-784-2960 or Markus H. Heim, Biomedicine Department, University of Basel, Hebelstrasse 20, 4053 Basel, Switzerland. E-mail: markus.heim@unibas.ch; fax: +41 61 265 38 47.

    Search for more papers by this author
  • Zuzanna Makowska,

    1. Department of Biomedicine, University of Basel, Basel, Switzerland
    Search for more papers by this author
  • Benedetta Campana,

    1. Department of Biomedicine, University of Basel, Basel, Switzerland
    2. Division of Gastroenterology and Hepatology, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
  • Diego Calabrese,

    1. Department of Biomedicine, University of Basel, Basel, Switzerland
    Search for more papers by this author
  • Michael T. Dill,

    1. Department of Biomedicine, University of Basel, Basel, Switzerland
    2. Division of Gastroenterology and Hepatology, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
  • Josan Chung,

    1. Department of Immunology and Microbial Science, Scripps Research Institute, La Jolla, CA
    Search for more papers by this author
  • Francis V. Chisari,

    Corresponding author
    1. Department of Immunology and Microbial Science, Scripps Research Institute, La Jolla, CA
    • Address reprint requests to: Stefan Wieland or Francis V. Chisari, Department of Molecular and Experimental Medicine, Scripps Research Institute, 10550, North Torrey Pines Road, La Jolla, CA 92037. E-mail: swieland@scripps.edu or fchisari@scripps.edu; fax: 858-784-2960 or Markus H. Heim, Biomedicine Department, University of Basel, Hebelstrasse 20, 4053 Basel, Switzerland. E-mail: markus.heim@unibas.ch; fax: +41 61 265 38 47.

    Search for more papers by this author
  • Markus H. Heim

    Corresponding author
    1. Department of Immunology and Microbial Science, Scripps Research Institute, La Jolla, CA
    2. Department of Biomedicine, University of Basel, Basel, Switzerland
    3. Division of Gastroenterology and Hepatology, University Hospital Basel, Basel, Switzerland
    • Address reprint requests to: Stefan Wieland or Francis V. Chisari, Department of Molecular and Experimental Medicine, Scripps Research Institute, 10550, North Torrey Pines Road, La Jolla, CA 92037. E-mail: swieland@scripps.edu or fchisari@scripps.edu; fax: 858-784-2960 or Markus H. Heim, Biomedicine Department, University of Basel, Hebelstrasse 20, 4053 Basel, Switzerland. E-mail: markus.heim@unibas.ch; fax: +41 61 265 38 47.

    Search for more papers by this author

  • See Editorial on Page 2065

  • Potential conflict of interest: Dr. Chisari advises and owns stock in Gilead. He consults and owns stock in Isis. He owns stock in Regulus, GlobeImmune, and Inovio.

  • This work reflects a collaboration that was initiated during M.H.H.'s sabbatical in F.V.C's laboratory at the Scripps Research Institute and continued in both laboratories thereafter. It was supported by NIH grants R01-AI079043 and U19-AI088778 (Project 2) to F.V.C, and Swiss National Science Foundation (SNF) grant 320030_130243 to M.H.H.

Abstract

Approximately 50% of patients with chronic hepatitis C (CHC) have ongoing expression of interferon stimulated genes (ISGs) in the liver. It is unclear why this endogenous antiviral response is inefficient in eradicating the infection. Several viral escape strategies have been identified in vitro, including inhibition of interferon (IFN) induction and ISG messenger RNA (mRNA) translation. The in vivo relevance of these mechanisms is unknown, because reliable methods to identify hepatitis C virus (HCV)-infected cells in human liver are lacking. We developed a highly sensitive in situ hybridization (ISH) system capable of HCV RNA and ISG mRNA detection in human liver biopsies and applied it to study the interaction of HCV with the endogenous IFN system. We simultaneously monitored HCV RNA and ISG mRNA using HCV isolate- and ISG mRNA-specific probes in liver biopsy sections from 18 CHC patients. The signals were quantified at the single-cell resolution in a series of random high-power fields. The proportion of infected hepatocytes ranged from 1%-54% and correlated with viral load, but not with HCV genotype or ISG expression. Infected cells occurred in clusters, pointing to cell-to-cell spread as the predominant mode of HCV transmission. ISG mRNAs were readily detected in HCV-infected cells, challenging previously proposed mechanisms of viral interference with the immune system. Conversely, infected cells and neighboring cells showed increased ISG mRNA levels, demonstrating that the stimulus driving ISG expression originates from HCV-infected hepatocytes. Conclusion: HCV infection in human hepatocytes during CHC does not efficiently interfere with IFN induction, IFN signaling, or transcription of ISG mRNA. (Hepatology 2014;59:2121–2130)

Abbreviations
CHC

chronic hepatitis C

IFN

interferon

ISG

interferon-stimulated gene

ISH

in situ hybridization

LP-AP

label probe conjugated to alkaline phosphatase

MAVS

mitochondrial antiviral signaling protein

OCT

optimal cutting temperature

pDC

plasmacytoid dendritic cell

peg

pegylated

pSVR

probability of sustained virological response

qPCR

quantitative polymerase chain reaction

RT

reverse transcription

Hepatitis C virus (HCV), a member of the flaviviridae, is a noncytopathic, positive-stranded RNA virus whose primary target cell is the hepatocyte.[1] To establish persistent infections, HCV has to evade and/or counteract the immune response of the host. In cell culture experiments, the viral protease NS3/4A has been shown to block interferon (IFN) regulatory factor 3 (IRF3) mediated transcriptional activation of IFNβ[2, 3] by cleavage and inactivation of mitochondrial antiviral signaling protein (MAVS)[4] and TIR-domain-containing adapter-inducing IFNβ (TRIF).[5] However, in vivo the innate immune system reacts within days after infection with the expression of hundreds of IFN-stimulated genes (ISGs).[6-8] Despite activation of the endogenous IFN system and the subsequent recruitment of HCV-specific T-cells to the liver, HCV persists in the majority of infected individuals. Once the infection progresses to the chronic phase, the number of infiltrating T-cells is reduced, and the predominantly IFNγ-driven gene expression found in the T-cell phase of acute hepatitis C is lost.[9] In about half of Caucasian patients, and in all chronically infected chimpanzees, hundreds of type I or type III IFN-stimulated genes are strongly induced in the chronic phase of hepatitis C, without having an evident antiviral effect on HCV.[7, 10-12] Moreover, patients with such an ongoing ISG expression, designated “preactivated” patients because their endogenous IFN system is already activated before therapy, have a poor response to treatment with pegylated IFN-α (pegIFN-α) and ribavirin.[11-13] The factors that determine the activation level of the endogenous IFN system in the liver include the IFNλ3 (IL28B) genotype,[14-16] the viral genotype[12] and the extent of MAVS cleavage in the liver.[17] It remains unclear why the endogenously induced expression of ISGs is ineffective against HCV, because the same set of genes is highly effective when induced in non-preactivated patients during treatment with pegylated IFNα and ribavirin.[12] It is conceivable that in the case of patients with a persistent activation of the endogenous IFN system in the liver, ISGs are expressed only in uninfected hepatocytes. This hypothesis implies that HCV interferes with IFN signaling through the Jak-STAT pathway and thereby blocks ISG expression in infected hepatocytes. There is some evidence from cell lines with ectopic expression of HCV proteins and from transgenic mice to support HCV interference with Jak-STAT signaling,[18, 19] albeit IFN signaling was not inhibited by HCV JFH-1 infection in Huh7.5 cells in vitro.[3] Alternatively, HCV could block the translation of ISG mRNAs at the ribosome, and there is strong support for this mechanism from cell culture experiments.[20] However, to clarify the relative importance of various viral interference strategies, in situ methods with single-cell resolution and reliable detection of HCV and of ISG products in liver biopsies of patients with chronic hepatitis C (CHC) are required.

Contrary to hepatitis B virus (HBV) infection, where core and surface antigens are readily detected in formalin-fixed, paraffin-embedded liver biopsies by routine immune-histochemistry,[21] detection of HCV proteins in human liver has only been sporadically reported and involved samples with very high viral load or required advanced imaging techniques such as two photon microscopy that are not widely available.[22] In situ detection of HCV RNA in liver biopsies was reported soon after the initial cloning of HCV,[23, 24] but those methods did not gain widespread acceptance. Here we demonstrate highly specific and sensitive detection of HCV infection at the cellular level by in situ hybridization (ISH) and apply this system to gain insight into the interaction between HCV and the innate immune response in the infected human liver.

Patients and Methods

Patients

HCV RNA and ISG mRNA ISH studies were performed with fresh-frozen liver biopsies from 18 patients (Table 1). The samples were selected from an existing biobank at the University Hospital Basel, Switzerland, to include different combinations of HCV genotype, viral load, and extent of hepatic ISG induction. HCV genotype and viral load were obtained from routine laboratory results generated during the clinical work-up of patients referred between October 2009 and April 2012 to the Hepatology Outpatient Clinic of the University Hospital Basel. Hepatic ISG induction was assessed using a previously published four-gene classifier that is based on a random forest classifier using the expression values of IFI27, RSAD2, ISG15, HTATIP2.[14] As controls for the ISH experiments, biopsies of one patient with chronic hepatitis B and three patients with alcoholic liver disease were included. All of them were negative for HCV. The biopsies for this study were embedded in TissueTek optimal cutting temperature (OCT) compound (Sakura, Horgen, Switzerland), immediately snap-frozen, and stored at −80°C until further processing. For RNA extraction an additional biopsy cylinder was snap-frozen and stored in liquid nitrogen.

Table 1. Clinical Characteristics of the 18 Patients Used for the HCV RNA Quantification by ISH on Fresh-Frozen Liver Biopsy Specimens
Biopsy NumberHCV GenotypeSerum Viral LoadLiver Viral Load% of HCV-Positive CellspSVRMetavir Grade and Stage
  1. Serum viral load is expressed as log10 IU HCV RNA/ml serum. Liver viral load is expressed as log10 HCV genome equivalents/μg total RNA in liver biopsy extracts. Percentage of HCV-positive cells was calculated based on the number of cells harboring at least one signal dot of HCV RNA staining in the single-color detection divided by the total number of cells counted. The Spearman correlation coefficient for the association of the HCV load with the percentage of HCV-positive cells was 0.7 for the serum viral load (P = 0.0015) and 0.6 for the liver viral load (P = 0.01). Probability of sustained virological response (pSVR) was calculated using the 4-gene classifier as described previously.[20] Biopsies were graded and staged according to Metavir classification (A0-4 for inflammatory activity grade, F0-4 for fibrosis stage).[33]

B7781a4.844.231.37%0.976A1/F0
B3211a4.944.113.21%0.16A2/F2
B8271a5.444.626.07%0A1/F1
B2241a5.744.195.43%0.080A1/F2
B7491a6.575.5032.33%0.196A1/F2
B2711a6.764.4434.44%0.982A3/F1
B2821b4.124.3114.00%0.996A1/F4
B8071b5.725.2535.12%0.382A3/F4
B2611b6.064.5539.36%0.646A1/F2
B3331b6.355.3625.37%0.002A3/F2
B29934.974.393.65%0.996A1/F2
B22135.853.649.88%0.478A2/F2
B22935.925.5253.97%0.988A2/F3
B24536.394.3133.65%0.998A2/F3
B63136.455.3421.77%0.418A2/F2
B31644.061.88%0.966A1/F1
B33945.313.9714.88%0.392A1/F2
B65946.515.2928.24%0.554A2/F1

For the analysis of probability of sustained virological response (pSVR) association with serum viral load we used data from 246 patients with CHC who had a liver biopsy at the University Hospital Basel between November 2005 and December 2012. Patient characteristics are included in Supporting Table 1.

All patients gave written informed consent to participate in this study, which was approved by the Ethics Committee of the Kantons Basel and Baselland.

HCV RNA Quantification in Liver Tissue and Sequencing for Probe Set Design

Total RNA from snap-frozen liver biopsy cylinders was extracted by the acid-guanidinium phenol-chloroform method.[25] Total liver RNA was DNase treated using the DNA-free system (Life Technologies, Grand Island, NY) per the manufacturer's instructions and 0.5 μg RNA were subjected to reverse transcription (RT) quantitative polymerase chain reaction (qPCR) as previously described.[26] Primer sequences used for RT-qPCR were HCV 5′-TCTGCGGAACCGGTGAGTA-3′ (sense) and 5′-TCAGGCAGTACCACAAGGC-3′ (antisense) targeting HCV sequences conserved in all HCV genotypes analyzed. HCV RNA levels were determined relative to standard curves comprised of serial dilutions of plasmids.

Patient-specific HCV target sequences for probe set design were determined by bulk sequencing of ∼2.2 kb of the HCV RNA genome spanning a very conserved region in the HCV genome[27] covering approximately nucleotide positions 4,000 to 6,200 including portions of NS3 and the NS4A and NS4B coding regions as depicted in Supporting Fig. 1. To do so, 0.5 μg total liver RNA was subjected to reverse transcription with Superscript III (Invitrogen, Carlsbad, CA) in 20 μL reactions containing 250 nM HCV genotype specific RT-primers followed by RNaseH (NEB, Ipswich, MA) digestion per the manufacturer's instructions. Two μL complementary DNA (cDNA) was then subjected to PCR amplification using primer sets (200 nM primer each) comprised of the HCV genotype-specific RT-primer and a corresponding primer about 2.2 kb upstream of the RT-primer location. PCR was performed in 50 μL reactions using the Phusion High-Fidelity DNA polymerase (NEB) per the manufacturer's instructions. PCR products were separated by gel electrophoresis and the 2.2 kb PCR product was isolated using the Qiagen gel extraction kit (Qiagen, Valencia, CA) and bulk sequenced (GeneWiz, San Diego, CA). In cases where this procedure did not produce the expected PCR product, reverse transcription was performed with random primers (hexamers) and overlapping 1 kb PCR fragments were produced to cover the entire probe set target region. Patient-specific HCV target sequences were assembled and analyzed using the Lasergene software package (DNASTAR, Madison, WI) and submitted to Panomics (Affymetrix, Santa Clara, CA) for probe set design and synthesis. HBV and IFI27-specific probe sets were produced by Panomics based on GenBank entries V01460 and NM_005532, respectively.

RNA ISH Staining Procedure

For the present study, we adapted a highly sensitive and specific ISH system (QuantiGene ViewRNA, Affymetrix) using HCV isolate specific probes.[28] Briefly, OCT-embedded liver biopsies were cryosectioned (8-9 μm thickness) in a cryostat, mounted onto Superfrost Plus Gold glass slides (Thermo Fischer Scientific, Wohlen, Switzerland), and kept at −80°C until use. For all experiments a negative control (liver biopsy section of an uninfected patient) was mounted onto the same slide and hybridized with the same conditions. Upon fixation (4% formaldehyde, 16-18 hours at 4°C), washing, and dehydration in ethanol, sections were pretreated by boiling for 1 minute in Pretreatment Solution, followed by 10 minutes digestion in Protease QF (both from Affymetrix). Sections were hybridized for 2 hours at 40°C with custom-designed QuantiGene ViewRNA probes against HCV RNA sequence targeting the positive HCV strand of each individual patient. In the duplex detection experiments, probes against IFI27 mRNA were added to the hybridization mixture. Bound probes were preamplified and subsequently amplified according to the manufacturer's instructions. Label Probe oligonucleotides conjugated to alkaline phosphatase (LP-AP) type 1 or type 6 were added, followed by the addition of Fast Red or Fast Blue Substrate used to detect HCV RNA. For duplex detection, an LP-AP type 6 probe was used with Fast Blue Substrate for IFI27 detection, followed by LP-AP type 1 probe with Fast Red Substrate for HCV detection. Finally, slides were counterstained with Meyer's hematoxylin and embedded with DAPI-containing aqueous mounting medium (Roti-Mount FluorCare DAPI, Roth, Arlesheim, Switzerland).

Image Acquisition

The images were acquired using a laser scanning confocal microscope (LSM710, Carl Zeiss Microscopy, Göttingen, Germany) and Zen2 software (Carl Zeiss Microscopy). For each liver biopsy section we initially acquired a nuclear stain image of the entire tissue cylinder (Supporting Fig. 2). Then, according to tissue area, we selected 15 to 25 nonoverlapping positions for high-power image acquisition based solely on the DAPI image. We avoided visible areas of high lymphocyte concentration and any other areas where the quality of DAPI staining would not allow reliable identification of nuclei. The high-power images (212.3 × 212.3 μm) were acquired using the 40× objective. Colors of each fluorescent dye were assigned during acquisition (red for Fast Red substrate, green for Fast Blue substrate, and blue for DAPI). The pictures were saved in Zeiss confocal file format (LSM), including multicolor layers.

Image Quantification and Data Analysis

Color channels of the acquired LSM images were separated and exported to TIF files using ImageJ software v. 1.43u.[29] The resulting files were then analyzed with CellProfiler software v. 2.0.[30] Nuclei were identified using DAPI staining and cell borders were approximated based on expanding the nuclear borders (Supporting Fig. 2). Red dots, corresponding to positive staining for HCV RNA, were identified based on round shape, specific size, and high staining intensity. The settings for HCV RNA signal identification were individually adjusted for each patient to allow for optimal quantification. For each single image we exported the original staining with overlaid outlines of the identified signals and all images were subsequently carefully reviewed to ensure that the signal identification process was correct. Finally, for each cell we exported the integrated intensity of the signal from the red dots and the number of red dots per cell, as well as the identities of the neighboring cells. For the 2-color staining we additionally exported the integrated intensity of the IFI27 staining in the green channel for each cell. Due to a more diffuse pattern of staining it was not possible to reliably identify and count the green dots of IFI27 mRNA. The count, intensity, and neighbor data exported from CellProfiler were analyzed with R statistical software.[31]

Results

Sensitive and Specific Detection of HCV RNA in Liver Biopsies From CHC Patients

We adapted a highly sensitive and specific ISH system (QuantiGene ViewRNA, Affymetrix) to detect HCV in liver biopsy tissue using isolate specific probes. The system is based on the simultaneous binding of multiple adjacent probe set pairs to the target RNA. Achieving high sensitivity and signal-to-noise ratio requires probe sets containing at least 20 highly specific probe set pairs covering ∼1 kb of target sequence. The HCV genome, however, does not contain a long enough region that is highly conserved between or within HCV genotypes suitable for probe set design. Thus, HCV plus strand-specific probe sets were designed and synthesized based on patient-specific consensus sequence information from an ∼2 kb fragment that spans the NS3, NS4A, and NS4B coding region that has been shown to be very conserved within a given patient HCV isolate.[27] Snap-frozen liver biopsies from 18 patients with CHC were analyzed (Table 1). The samples were selected from the preexisting biobank based on a combination of HCV genotype, viral load, and extent of hepatic ISG induction. For each sample, HCV RNA was isolated and sequenced from the same liver biopsy that was used for ISH procedures. Using this method we were able to reliably detect HCV RNA in the liver of patients with viral loads as low as 104 units/mL (Table 1). In all samples, HCV-specific signals were detected exclusively in hepatocytes (Fig. 1; Supporting Fig. 3A). Importantly, this system was characterized by an extremely high signal-to-noise ratio with virtually no unspecific background staining and no crossreactivity by probe sets targeting plus or minus strand HCV RNA (Supporting Fig. 3B), HBV or HCV viral RNAs (Supporting Fig. 4), and even between HCV isolates with identical genotypes, we found only weak crossreactivity (Supporting Fig. 5). However, achieving the high sensitivity and signal to noise ratio required probe sets containing at least 20 probe pairs covering ∼1 kb of the HCV genome.

Figure 1.

In situ HCV RNA detection in human liver tissue across a range of serum viral loads. (A-C) Single-color HCV RNA detection by ISH in patients with high (A, 5,754,399 IU/mL), middle (B, 707,946 IU/mL), and low (C, 87,096 IU/mL) serum viral load. (D) Liver sample from an HCV-negative patient (−CT) stained with the probe set directed against the B271 HCV isolate. Biopsy numbers are shown in the top right corners. Scale bars = 25 μm.

Percentage of Infected Hepatocytes Determines Serum Viral Load

In each biopsy, the number of HCV-infected hepatocytes was assessed in 15-25 randomly chosen quadrants of 0.04 mm2 (Supporting Fig. 2). The percentage of infected hepatocytes in the 18 biopsies ranged from 1.3% to 53.9% (Table 1, Fig. 2A). There was a significant positive correlation between the proportion of infected hepatocytes (phcv) and the viral load in the serum (Fig. 2A,B) and the liver (Supporting Fig. 6), but not with the HCV genotype (Supporting Fig. 7). The number of HCV signal dots per infected hepatocyte varied considerably from 1 to 25, but in most infected hepatocytes it did not exceed 2 (Supporting Fig. 8, Supporting Table 2). The average number of HCV signal dots per infected hepatocyte (Nhcv) significantly correlated with phcv (Fig. 2C) and also correlated with serum viral load. However, in a multiple linear regression analysis, Nhcv did not significantly improve the predictive power of phcv for serum viral load. We conclude that serum viral loads are predominantly determined by the number of infected cells. HCV-positive cells occurred in clusters. The probability of a given cell to have a HCV-positive neighbor cell significantly depended on its own infection status. Infected cells have infected neighbors twice as often as uninfected cells (Fig. 2D), and the probability of having infected neighbors significantly increases further in cells with more than two HCV signal dots. In agreement with Liang et al.,[22] we conclude that in an infected liver, HCV spreads predominantly locally, possibly through direct cell-to-cell transmission without an extracellular phase. Such an infection modus would have the advantage of not being subject to inhibition by neutralizing antibodies.

Figure 2.

HCV-infected cells occur in clusters in human liver tissue and their numbers correlate with serum viral load. (A) Boxplot showing the percentage of liver cells positive for HCV RNA in 18 patients. Patients with high ISG expression in the liver (pSVR ≤ 0.5) are shown in red and patients with low ISG expression (pSVR > 0.5) in blue. Percentage of HCV-positive cells was calculated based on the number of cells harboring at least one signal dot of HCV RNA staining in the single-color detection divided by the total number of cells counted. Each dot represents the proportion of infected cells in one randomly chosen image. For every patient, data from one to three independently stained liver biopsy sections are pooled. The blue line shows log10 serum viral load expressed in IU/mL. (B) Percentage of liver cells positive for HCV RNA is associated with serum viral load. Association between the variables was evaluated using Spearman's rank correlation coefficient (rho). Patients with high ISG expression in the liver (pSVR ≤ 0.5) are shown in red and patients with low ISG expression (pSVR > 0.5) in blue. (C) Percentage of liver cells positive for HCV RNA is associated with the mean number of HCV signal dots per infected cell. Association between the variables was evaluated using Spearman's rank correlation coefficient (rho). Patients with high ISG expression in the liver (pSVR ≤ 0.5) are shown in red and patients with low ISG expression (pSVR > 0.5) in blue. (D) Infected cells are more likely to have infected neighbors than noninfected cells. Each dot shows the proportion of cells that have at least one HCV-positive adjacent neighbor. Lines connect measurements from the same patient. HCV−: noninfected cells; HCV+: infected cells; HCV++: infected cells which harbor more than two HCV signal dots. P values were calculated using paired two-tailed Student t test.

Frequency of HCV-Positive Cells in the Liver is Independent of the Level of Intrahepatic Interferon-Stimulated Gene Expression

It is now well established that patients with constitutive high expression levels of ISGs in the liver are poor responders to pegylated IFNα and ribavirin.[11-14] We have previously developed a random forest classifier that predicts the pSVR based on the expression level of 4 genes (IFI27, ISG15, RSAD2, and HTATIP2) in the liver.[14] In the CHC patient population followed at the University Hospital Basel, the frequency distribution of pSVR is U-shaped (Fig. 3A). Of note, the extent of ISG expression (quantitatively assessed with the 4-gene classifier and expressed as pSVR) does not correlate with serum viral load (Fig. 3B). To investigate the spatial relationships of HCV and the host innate immune system we adapted a multiplex fluorescence ISH (FISH) staining for use in human liver biopsies. We could simultaneously detect HCV RNA and mRNA of IFN-stimulated genes like IFI27 (Fig. 3C-F) or USP18 (data not shown). Importantly, the mean value of IFI27 integrated intensity per cell correlated with the amount of IFI27 mRNA relative to GAPDH measured by qPCR in the same biopsies (Supporting Fig. 9A,B). Also, there was a very good concordance of the percentages of HCV-positive cells obtained using the one-color and multiplex approaches, further validating the results obtained using the multiplex detection system (Supporting Fig. 9C,D). Multiplex ISH experiments with simultaneous detection of HCV RNA and ISG mRNAs revealed no correlation between the number of cells expressing ISG mRNAs (IFI27 or USP18) and the number of HCV-infected cells (data not shown). We had samples with extensive HCV staining but little IFI27 expression (Fig. 3C), samples with extensive HCV staining and strong IFI27 expression (Fig. 3D), samples with little HCV and IFI27 signals (Fig. 3E), and samples with little HCV but extensive IFI27 signals (Fig. 3F). These findings are in agreement with the lack of a correlation between serum viral load and the extent of ISG expression assessed by the 4-gene classifier (Fig. 3B). Apparently, viral replication cannot be efficiently restricted by the up-regulation of hundreds of ISGs observed in a large proportion of patients with CHC.

Figure 3.

Extent of ISG induction in HCV patients does not correlate with serum viral load or percentage of infected cells. (A) Distribution of SVR probabilities calculated according to the four-gene classifier[14] in a sample of 246 CHC patients. (B) There is no association of serum viral load with probability of SVR calculated according to the four-gene classifier in a sample of 221 CHC patients (for 25 patients with known pSVR the serum viral load was not available). (C-F) Representative examples of two-color ISH staining for HCV RNA (in red) and IFI27 mRNA (in green) in patients with: (C) low expression levels of IFI27 (0.084 relative to GAPDH) and high viral load (831,762 IU/mL); (D) high expression levels of IFI27 (1.8 relative to GAPDH) and high viral load (2,818,383 IU/mL); (E) low expression levels of IFI27 (0.28 relative to GAPDH) and low viral load (46'773 IU/mL); (F) high expression levels of IFI27 (2.3 relative to GAPDH) and low viral load (87,096 IU/mL). Biopsy numbers are shown in the top right corners. Scale bars = 25 μm.

HCV-Infected Hepatocytes and the Neighboring Cells Show Increased Levels of ISG Expression

To further explore potential mechanisms of ISG induction in CHC, we performed a quantitative analysis of the spatial relationship between HCV RNA and ISG mRNA expression in the subset of patients with an induced endogenous IFN system. We observed that the signal intensity of IFI27 staining was lowest in uninfected cells with uninfected neighbors, intermediate in uninfected cells with at least one HCV-positive neighbor, and highest in HCV-positive cells (Fig. 4A,B). We observed the same pattern of IFI27 staining in samples from patients without an induced endogenous IFN system, but in these samples the IFI27 staining intensity was lower. These results demonstrate that in the majority of HCV-infected hepatocytes Jak-STAT signaling and ISG transcription are not inhibited (Fig. 4A-C). Occasionally, high levels of ISG mRNA could also be found in uninfected cells without any HCV-positive neighbors, pointing to the possibility of long-range IFN signals (Fig. 4D). A careful inspection of the colocalization of HCV RNA and ISG mRNA signals rarely identified HCV-positive cells with little or no ISG transcription and ISG-negative neighbors (Fig. 4E) and HCV-positive cells without ISG expression but with neighbors that strongly express ISGs (Fig. 4F).

Figure 4.

HCV-infected cells and their neighbors have increased probability of high ISG expression compared to noninfected cells. (A) Mean integrated intensity of IFI27 mRNA ISH staining per cell in the six patients with high expression of hepatic ISGs (pSVR ≤ 0.5) for whom double-color staining was available (B631, B749, B339, B333, B321, and B827). Every cell from each patient was grouped into one of the three categories: 1) uninfected cells without any infected adjacent neighbors; 2) uninfected cells with at least one infected adjacent neighbor; 3) infected cells; and the average integrated intensity of IFI27 mRNA staining per cell in each category was calculated. Lines connect measurements from the same patient. P values were calculated using paired two-tailed Student t test. (B) Integrated intensity of IFI27 mRNA ISH staining per cell in a selected patient with induced ISG expression (B749). Each cell was grouped into one of the three categories: 1) uninfected cells without any infected adjacent neighbors; 2) uninfected cells with at least one infected adjacent neighbor; 3) infected cells. P values were calculated using two-tailed Student t test. (C-F) Selected images of two-color ISH staining for HCV RNA (in red) and IFI27 (green) mRNA in the liver biopsy from one representative patient (B749) with induced ISGs. The arrows indicate examples of different HCV-ISG localization patterns: (C) colocalization of HCV RNA signal dots with high levels of IFI27 mRNA in the same cell; (D) high levels of IFI27 mRNA despite the lack of HCV RNA-positive cells in direct proximity; (E) very weak IFI27 signals in HCV-infected cells and neighboring noninfected cells; (F) very weak IFI27 signals in an HCV-infected cell and strong IFI27 signals in surrounding cells (both HCV-negative and HCV-positive). Scale bars = 25 μm.

Discussion

We have developed a highly sensitive, specific, and reproducible detection system for HCV RNA in liver biopsies of patients with CHC and viral loads as low as 104 IU/mL. This methodological breakthrough allowed us to address heretofore unapproachable fundamental questions about host-virus interactions in CHC. Regardless of HCV genotype, serum viral load positively correlates with the proportion of infected hepatocytes. Patients with less than 105 IU/mL HCV RNA in the serum have fewer than 5% of their hepatocytes infected. In patients with high viral load, the proportion of infected hepatocytes is in the range of 20%-50%. These results have important implications for the interpretation of results obtained from the analysis of liver biopsies of patients with CHC. In biopsy samples from patients with serum viral loads below 105 IU/mL any biochemical changes present in the infected cells will be diluted by 95% of uninfected cells. Consequently, the magnitude of changes of parameters that are supposedly directly regulated by HCV in a cell-autonomous way, such as, for example, dysregulation of the intermediate metabolism, rearrangement of cellular compartments, or adaptive mechanisms such as endoplasmic stress response, will depend on the number of infected hepatocytes. Our present results showing a strong correlation of serum viral load with the percentage of infected hepatocytes provide a rationale for selecting biopsy samples from patients with high serum viral load for the detection of HCV induced cell-autonomous changes.

Both the proportion of infected hepatocytes and the serum viral load does not depend on the activation level of the endogenous IFN system in the liver, and does not correlate with IFNλ3 (IL28B) genotype. HCV-infected cells appear in clusters, an observation that is in agreement with data from two-photon microscopy analysis of HCV protein expression[22] and HCV RNA quantification in groups of 100-200 hepatocytes obtained by laser capture microdissection of liver biopsies,[32] and is reminiscent of immunohistochemical findings in chronic hepatitis B.[21] The clustered spatial distribution of HCV-infected cells suggests that the predominant mode of infection in the liver is by cell-to-cell spread of HCV (with or without an extracellular phase of the HCV life cycle).

In samples from patients with an induced endogenous IFN system, ISGs were highly expressed in most of the HCV-infected hepatocytes. These results do not support a relevant role of MAVS cleavage as an escape strategy of HCV in preactivated patients during the chronic phase of the infection, but rather point to HCV-infected hepatocytes as the major source of IFN production, either cell autonomously or through cell-cell interaction with IFN producing nonparenchymal cells, e.g., plasmacytoid dendritic cells (pDCs).[28, 33] Furthermore, the finding also excludes that HCV interference with Jak-STAT signaling or with ISG transcription is a major viral escape strategy. We cannot formally exclude that in a minority of cells, HCV can inhibit the induction of ISGs, because a careful inspection of the images revealed rare cells that were HCV-positive but had little or no ISG transcription (Fig. 4E). Along the same line of reasoning, the rare detection of HCV-positive cells without ISG expression but with neighbors that strongly express ISGs (Fig. 4F) could be due to efficient HCV interference with IFN signal transduction in HCV-infected cells. However, both situations were uncommon in the biopsies of all patients with induced ISG expression. In accordance with results from studies in HCV-infected Huh7.5 cells[3] we have to conclude that, in general, HCV neither prevents the induction of IFN nor signal transduction through the Jak-STAT pathway. Since our in vivo results do not support either of these two hypotheses, by default they suggest that another mechanism is more likely to explain the ability of HCV to persist in the face of a robust ISG mRNA response, i.e., HCV interference with ISG mRNA translation as previously suggested by Garaigorta and Chisari.[20] However, formal proof of this hypothesis requires simultaneous single-cell analysis of intracellular HCV RNA, ISG mRNA, and ISG protein in the biopsies, which is technically not feasible at the moment.

In conclusion, the spatial correlation of HCV RNA signals with ISG mRNA expression in patients with induced ISG expression reveals that HCV is the central driver of ISG induction. This finding is in agreement with the positive correlation of HCV viral load and hepatic type I ISG expression that was found during the acute phase of HCV infection in chimpanzees.[7, 8] Despite extensive attempts to detect IFNα, IFNβ, or IFNλ mRNA with this highly sensitive ISH method, we could not identify the type of IFN or the cellular source of the IFN(s) that drive ISG expression in CHC patients. However, we are confident to conclude that the original stimulus that results in ISG expression comes from HCV-infected hepatocytes.

Acknowledgment

We thank Christina Whitten-Bauer and Sylvia Ketterer for excellent technical assistance. This is manuscript # 27002 from The Scripps Research Institute.

Ancillary