Protein modification with ISG15 blocks coxsackievirus pathology by antiviral and metabolic reprogramming

Abstract

Protein modification with ISG15 (ISGylation) represents a major type I IFN-induced antimicrobial system. Common mechanisms of action and species-specific aspects of ISGylation, however, are still ill defined and controversial. We used a multiphasic coxsackievirus B3 (CV) infection model with a first wave resulting in hepatic injury of the liver, followed by a second wave culminating in cardiac damage. This study shows that ISGylation sets nonhematopoietic cells into a resistant state, being indispensable for CV control, which is accomplished by synergistic activity of ISG15 on antiviral IFIT1/3 proteins. Concurrent with altered energy demands, ISG15 also adapts liver metabolism during infection. Shotgun proteomics, in combination with metabolic network modeling, revealed that ISG15 increases the oxidative capacity and promotes gluconeogenesis in liver cells. Cells lacking the activity of the ISG15-specific protease USP18 exhibit increased resistance to clinically relevant CV strains, therefore suggesting that stabilizing ISGylation by inhibiting USP18 could be exploited for CV-associated human pathologies.

Fig. 1. ISGylation suppresses CV titers at different phases of infection.

Wild-type (wt), ISG15^−/−, and Ube1L^−/− mice were infected with 1 × 10⁵ pfu of CV Nancy and sacrificed at the indicated points in time post infection (p.i.). Tissue from (A) liver and (B) heart of wild-type mice was subjected to Western blot analysis using an ISG15-specific antibody. Each lane represents tissue homogenates obtained from a different animal. (C and D) Infectious viral particles were quantified in the respective organs obtained from wild-type, ISG15^−/−, and Ube1L^−/− mice by plaque assay during CV infection. Each dot represents a different animal; data are summarized as median values. Student’s t tests were conducted. P values of <0.05 are indicated in the graph. (E to J) USP18C61A and wild-type littermate controls were infected with 1 × 10⁵ pfu of CV and sacrificed at the indicated points in time. (E) Heart tissue was homogenized and subjected to Western blot analysis for detection of ISG15. Each lane represents tissue obtained from a different animal, and the shown example for each group and point in time is representative for n = 3 mice. (F) At day 6 after infection, infectious viral particles were determined in heart by TaqMan qPCR and plaque assay. Each dot represents a different animal. Data are summarized as means ± SEM; t tests were performed, and P values of <0.05 are depicted. (G) Cardiomyocytes derived from USP18C61A ISG15^−/− and ISG15^−/− embryos were transduced with Ad5 vectors encoding murine ISG15 and stimulated with poly(I:C). Cellular lysates were subjected to Western blot analysis. (H to J) ISG15-rescued cardiomyocytes from USP18C61A ISG15^−/− and ISG15^−/− embryos were infected with CV at an MOI of 0.1 for 24 hours. (H) Total RNA was isolated to determine CV genome copy numbers by TaqMan qPCR; ΔΔC_t values obtained from duplicates are shown for two independent experiments. Data are summarized as means ± SEM. (I) Viral load was determined in cellular lysates by Western blot analysis of CV VP1. Densitometric analysis of VP1 and the GAPDH loading control was performed to calculate the mean (± SEM) relative VP1 expression in three independent experiments. (J) Release of infectious viral particles was assessed by plaque assay. One representative out of three independent experiments yielding the same result is depicted. Unpaired t test (plaque assay) and one-sample t tests (VP1 and CV RNA) were performed, and P values are depicted.

Fig. 2. Protection from CV pathology requires ISGylation in nonhematopoietic cells.

(A and B) ISG15^−/− (CD45.2) mice were reconstituted with bone marrow cells from either ISG15^−/− (CD45.2) or wild-type (wt, CD45.1) donors before CV infection, and mice were sacrificed 3 days after infection. (A) Infectious viral particles were quantified by plaque assay (wild-type ➔ ISG15^−/−, n = 6; ISG15^−/− ➔ ISG15^−/−, n = 4). Data are summarized as median. (B) Splenic mRNA expression of the indicated cytokines and chemokines was determined by TaqMan qPCR. (C to F) Chimeric wild-type and Ube1L^−/− mice were generated upon transfer of wild-type or Ube1L^−/− bone marrow cells into lethally irradiated wild-type or Ube1L^−/− recipients, respectively. Mice were infected with CV and sacrificed after 8 days (n = 7 in all four groups). (C) Infectious viral particles were quantified in heart tissue by plaque assay. Data are summarized as means ± SEM. (D) Myocarditis was scored microscopically by a blinded pathologist based on cardiac hematoxylin and eosin staining. (E) Representative histopathologic stains of heart tissue of each group are shown. (F) mRNA levels of the indicated genes in heart tissue were determined by TaqMan qPCR. Unequal variance versions of two-way ANOVA were performed, followed by a Sidak-Holm’s multiple comparison test. Data were summarized as means ± SEM if applicable.

Fig. 3. ISGylation increases expression levels of IFIT1/3 proteins.

(A) Liver samples obtained from wild-type (wt) mice (n = 3) before and 3 days after infection were subjected to LC-MS/MS analysis. In the depicted volcano plots, each protein identified at baseline and after CV infection in the proteome screen of liver tissue is represented by a dot. The x axis specifies log₂ fold changes and the y axis specifies −log₁₀ of P values obtained from t test. (B) 3×FLAG-6×His-ISG15-HeLa cells and controls were stimulated with IFN-β for 24 hours, and cellular lysates were subjected to FLAG immunoprecipitation (IP). Total cell lysates (input) and FLAG-IP enriched proteins were immunoblotted using antibodies for ISG15, IFIT3, and IFIT1. Data are representative of three independent experiments. (C) HeLa cells transfected with a four-plasmid combination (HA-ISG15, Ube1L, Ube2L6, and Herc5) and with either FLAG-tagged IFIT1 or IFIT3 were subjected to pull-down experiments using anti-FLAG agarose beads. Total cell lysates and FLAG-immunoprecipitated proteins were subjected to immunoblotting using antibodies for FLAG and HA. Data are representative of three independent experiments. (D and E) ISG15-ko HeLa cells generated by CRISPR-Cas9–mediated gene editing and respective control cells were stimulated with IFN-β (100 U/ml). IFIT1/IFIT3 transcripts and IFIT1/IFIT3 protein expression levels were determined by (D) TaqMan qPCR and (E) Western blot analysis. One representative of three independently performed experiments is shown. Unpaired t tests were conducted. (F) ISG15-ko HeLa cells and controls were transfected with plasmids encoding either FLAG-tagged IFIT1 or IFIT3, and IFIT expression was determined by Western blot analysis. Date are representative of three independent experiments. (G) ISG15-ko HeLa cells were transfected with either GFP or ISG15 cDNA together with IFIT1 or IFIT3 expression plasmids, respectively. Subsequently, cells were stimulated with IFN-β for 24 hours, and IFIT protein expression levels were determined by Western blot analysis. Data are representative of two independent experiments. (H) Following the experimental approach outlined in (G), ISG15-ko cells were transfected with either ISG15 or the unconjugatable ISG15-LRAA mutant as well as with IFIT1/3 cDNA. IFIT protein expression was determined after 24 hours by Western blot analysis. Data are representative for three independent experiments. (I) Primary cardiomyocytes were generated from USP18C61A ISG15^−/− and ISG15^−/− mice, and ISGylation was induced by transduction of Ad5 vectors encoding mISG15. Cells were left unstimulated or stimulated with IFN-β (100 U/ml) to induce the expression of endogenous IFIT1 and IFIT3.

Fig. 4. ISG15 influences the metabolic state in healthy mice and during CV infection.

(A) Unstimulated or IFN-β pretreated, ISG15-ko HeLa, and control cells were infected with CV (MOI 1.0). After 32 hours, the glucose and lactate concentrations were determined in cell culture supernatants. Values normalized to cell counts per well are presented as increase relative to the untreated wild-type control. The data of four independent experiments have been summarized as means ± SEM. A two-way ANOVA test was performed followed by a Sidak’s multiple comparison test. (B to D) Wild-type (wt) and ISG15^−/− mice sacrificed before (wild-type, n = 7; ISG15^−/−, n = 7), 3 days (wild-type, n = 8; ISG15^−/−, n = 6), and 8 days (wild-type, n = 9; ISG15^−/−, n = 8) after CV infection. (B) Blood glucose and serum triglycerides/FFAs at baseline and 3 and 8 days after CV infection in wild-type and ISG15^−/− mice summarized as the means ± SEM. Numbers in bar graphs are the percentage of the metabolite compared to the mean value in naive wild-type mice. Multiple t tests with Holm-Sidak multiple comparison correction were performed. (C) Glycogen content visualized in liver tissue section with periodic acid Schiff reaction. Representative images are shown. (D) Glycogen content scored semiquantitatively from 0 to 4 based on histological analysis and glucose concentration assessed in liver tissue homogenates. Bars are means ± SEM. Multiple t tests with Holm-Sidak multiple comparison correction were performed; P values are depicted.

Fig. 5. ISG15 reprograms the central liver metabolism during CV infection.

(A) Hepatic tissue obtained from wild-type (wt) and ISG15^−/− mice (n = 3) during early (day 3) and late (day 8) state of CV infection was subjected to LC-MS/MS analysis. Heatmaps summarizing all differentially regulated hepatic proteins during infection for both strains are depicted. The relative abundance of each protein is color-coded based on the z score normalized log₂-transformed LFQ intensities. Blue color indicates proteins of high abundance, and yellow color indicates proteins of low abundance as compared to row means. A hierarchical clustering resolved six distinct clusters, with annotation shown on the right. (B) Heatmap-based clusters were subjected to Gene Ontology (GO) analysis, and proteins involved in selected enriched metabolic GO terms with catabolic ATP-generating function (FA oxidation, carbohydrate catabolic process, and OXPHOS) are depicted at an early and late state of CV infection, applying the same color code as used in (A) (blue, up-regulation; yellow, down-regulation). If the GO term of interest was not found within a dataset, individual proteins were not plotted. (C) At the indicated time points of infection, liver biopsies were obtained from wild-type and ISG15^−/− mice. The basal oxygen consumption (top) and extracellular acidification (bottom) rates were monitored using a Seahorse Biosciences extracellular flux analyzer. Values were normalized to protein content in the biopsies. Data of at least six mice per group were summarized as means ± SEM. A one-way ANOVA was performed followed by a Tukey’s multiple comparison test. (D and E) Liver proteome data together with HEPATOKIN1, a model of central liver metabolism (43), were used to assess the metabolic alterations in liver tissue of wild-type and ISG15^−/− mice during viral infection. Metabolic models for the different conditions were constructed by scaling the maximal activity for each enzyme using the LFQ intensities for each protein obtained from MaxQuant analysis at the respective point in time. (D) For a standard 24-hour profile metabolite plasma profile, diurnal glucose exchange fluxes were simulated in wild-type and ISG15^−/− mice at each time point of viral infection. Negative exchange fluxes indicate net release from the liver to the plasma (gluconeogenesis), while positive values indicate hepatic glucose uptake (glycolysis). (E) For each condition, experimentally determined blood glucose levels as depicted in Fig. 4B were used as model input to calculate realistic exchange fluxes and glycogen levels.

Fig. 6. Human ISG15 suppresses CV replication.

(A) ISG15 expression was deleted in HeLa cells using CRISPR-Cas9 gene editing as described in Fig. 3. ISG15-ko cells and wild-type cells were infected with CV (MOI 0.1) for 16 hours. Expression of CV VP1 was determined by Western blot analysis in four independent experiments, and the obtained signal normalized to GAPDH was compared to wild-type samples. Infectious virus particles were quantified in five independent experiments by plaque assay. Data are summarized as means ± SEM. (B) HeLa cells were transfected with siRNA targeting human ISG15 or a nontargeting control siRNA. Cells were subsequently infected with CV (MOI 0.01) for 16 hours. VP1 protein expression was determined and normalized to the control sample in four independent experiments. Plaque assays were performed in two independent experiments. Data are summarized as described in (A). (C) HeLa cells stably expressing FLAG-tagged human ISG15 and respective control cells were infected with CV (MOI 0.1) for 16 hours. VP1 expression and virus titer were determined as described in (A) in three independent experiments. (D) Primary embryonic cardiomyocytes obtained from ISG15^−/− mice were transduced with Ad5 vectors expressing human ISG15 (hISG15) or control for 48 hours at MOI 25 before CV infection (MOI 0.1) for 24 hours. VP1 levels were determined by Western blot analysis, and infectious viral particles were quantified by plaque assay in three independent experiments. (E) Cardiomyocytes derived from USP18C61A ISG15^−/− and ISG15^−/− embryos were transduced with Ad5 vectors encoding hISG15 and infected with CV in three independent experiments for detection of the viral load by Western blot analysis of VP1 as well as plaque assay. One-sample t tests were performed for all summarized VP1 data. Unpaired t tests were conducted for all plaque assay data. (F and G) CVB isolates were obtained from patients presenting with neurological symptoms that may have been of infectious origin. (F) ISG15-ko HeLa cells and control cells were infected with the indicated CV serotypes (MOI 0.1), and infectious virus particles were determined after 16 hours by plaque assay. The relative increase of the viral titer in ISG15-ko HeLa cells as compared to control cells is depicted for a representative experiment. Three independent experiments demonstrated similar results. (G) The ISG15 system was induced in USP18C61A ISG15^−/− and ISG15^−/− cardiomyocytes by transduction of Ad5 vectors encoding hISG15. Cells were infected with CV serotypes as follows: CVB1 425, MOI 1 (2 days); CVB1 506, MOI 1 (1 day); CVB3 1072, MOI 10 (2 days); CVB3 180, MOI 10 (1 day); CVB4 686, MOI 10 (2 days); CVB4 120, MOI1 (2 days); CVB5 679, MOI 1 (1 days); CVB5 800, MOI 10 (1 day). Infectious viral particles were quantified by plaque assay. Relative reduction of the viral load in USP18C61A ISG15^−/− cells as compared to ISG15^−/− cardiomyocytes with restored ISG15 expression is depicted for a representative of at least three independent experiments. Data are means ± SEM; one-sample t tests were performed, and P values are depicted.

Fig. 7. Protein modification with ISG15 acts cooperatively with IFIT proteins and preserves glucose homeostasis.

CV infection is a bona fide example for multiphasic state infectious disease with primary injury of liver and pancreas followed by a second viremia culminating in cardiac damage and chronic tissue damage. Early upon infection, IFNs trigger the ubiquitin-like modifier ISG15, which, in a three-step enzymatic cascade, forms covalent linkages with proteins in both infected and noninfected cells. In non–bone marrow–derived somatic cells and tissues, ISGylation inhibits viral replication, and this involves augmented protein expression levels of antiviral effectors such as IFIT1 and IFIT3. ISG15 ensures efficient storage of glucose in liver tissue of healthy mice and reprograms liver metabolism toward improved glucose production early after CV infection. Cells lacking activity of the ISG15-specific protease USP18 show a marked increased resistance against CV infection, thus providing a rationale that USP18 inhibition could be a novel host-directed approach countered to CV-associated human pathology.