The SARS-CoV-2 Nsp3 macrodomain reverses PARP9/DTX3L-dependent ADP-ribosylation induced by interferon signaling

Abstract

SARS-CoV-2 nonstructural protein 3 (Nsp3) contains a macrodomain that is essential for coronavirus pathogenesis and is thus an attractive target for drug development. This macrodomain is thought to counteract the host interferon (IFN) response, an important antiviral signalling cascade, via the reversal of protein ADP-ribosylation, a posttranslational modification catalyzed by host poly(ADP-ribose) polymerases (PARPs). However, the main cellular targets of the coronavirus macrodomain that mediate this effect are currently unknown. Here, we use a robust immunofluorescence-based assay to show that activation of the IFN response induces ADP-ribosylation of host proteins and that ectopic expression of the SARS-CoV-2 Nsp3 macrodomain reverses this modification in human cells. We further demonstrate that this assay can be used to screen for on-target and cell-active macrodomain inhibitors. This IFN-induced ADP-ribosylation is dependent on PARP9 and its binding partner DTX3L, but surprisingly the expression of the Nsp3 macrodomain or the deletion of either PARP9 or DTX3L does not impair IFN signaling or the induction of IFN-responsive genes. Our results suggest that PARP9/DTX3L-dependent ADP-ribosylation is a downstream effector of the host IFN response and that the cellular function of the SARS-CoV-2 Nsp3 macrodomain is to hydrolyze this end product of IFN signaling, rather than to suppress the IFN response itself.

Keywords: ADP-ribosylation; COVID-19; DTX3L; PARP9; SARS-CoV-2; macrodomain.

Figure 1

Both type I and type II IFN signaling induce ADP-ribosylation in A549 cells.A, immunoblot for STAT1 phospho-Y701 and actin loading control in A549 cells 24 h after treatment with either vehicle control, recombinant interferon gamma (IFNγ), or transfection with poly(I:C), at the doses shown. B, left, representative images of immunofluorescence staining for STAT1 phospho-Y701 in A549 cells 24 h after treatment with vehicle control, 100 U/ml IFNγ or transfection with 0.1 μg/ml poly(I:C); right, quantification of mean p-STAT1 fluorescence per nucleus, averaged for thousands of cells per replicate and normalized to the IFNγ-treated sample. Mean ± SEM (n = 4, from three separate experiments), ∗∗∗∗p < 0.0001. Scale bars = 20 μm. C, left, representative images of immunofluorescence staining for ADP-ribose modification (pan-ADP-ribose - Millipore) in A549 cells 24 h after treatment with vehicle control, 100 U/ml IFNγ or transfection with 0.1 μg/ml poly(I:C); right, quantification of total ADP-ribose fluorescence in cytosolic dots per cell, averaged for thousands of cells per replicate, and normalized to the IFNγ-treated sample. Mean ± SEM (n = 4, from three separate experiments), ∗∗∗∗p < 0.0001. Scale bars = 20 μm.

Figure 2

Ectopic expression of the Nsp3 macrodomain reverses IFN-induced ADP-ribosylation.A, representative images of anti-FLAG immunofluorescence staining (green) and DAPI staining (blue) in A549 cells transduced either with an empty vector control (left) or with a lentiviral construct constitutively expressing FLAG-tagged SARS-CoV-2 Nsp3 macrodomain (right). Scale bars = 20 μm. B, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain 24 h after treatment with vehicle control, 100 U/ml IFNγ or transfection with 0.1 μg/ml poly(I:C). Mean ± SEM (n = 8, from three separate experiments), ∗∗∗∗p < 0.0001. C, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells transduced either with empty vector control (e.v.) or with lentiviral constructs for constitutive expression of either WT macrodomain or catalytically dead N40A mutant, after 24 h treatment with 1000 U/ml IFNα, 1000 U/ml IFNβ, 100 U/ml IFNγ, transfected with 0.1 μg/ml poly(I:C) or vehicle control. For FLAG-macrodomain-expressing samples, cells were gated such that macrodomain expression between WT and N40A mutant was comparable (Fig. S2, A–C). Mean ± SEM (n = 6, from three separate experiments), ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001. D, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells transduced either with empty vector control (e.v.) or with lentiviral constructs for doxycycline-inducible expression of either WT macrodomain or catalytically dead N40A mutant, after 24 h treatment with indicated doses of doxycycline and 100 U/ml IFNγ or vehicle control. For FLAG-macrodomain-expressing samples, cells were gated such that macrodomain expression between WT and N40A mutant was comparable (Fig. S2, E–G). Mean ± SEM (n = 7–12, from four separate experiments), ∗∗∗p < 0.001 and ∗∗∗∗p < 0.0001.

Figure 3

A repurposing screen for macrodomain inhibitors validates the technique for further screening.A, virtual screening workflow, starting with 6365 compounds approved for human consumption by any regulatory agency in the world, reaching a final list of 79 compounds taken forward for testing, of which 69 were sourced. B, thermal shift assays, by nanoDSF, of the recombinant Nsp3 macrodomain in the absence (blue) or presence of 100 μM ADP-ribose (orange) or 100 μM of each of the 69 test compounds (black). Melting temperatures measured as the inflection point of the 350 nm/330 nm intrinsic fluorescence ratio, indicative of protein unfolding (top) and onset of light scattering, indicative of protein aggregation (bottom) are shown. C, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of WT macrodomain, 24 h after treatment with vehicle control, 100 U/ml IFNγ or 100 U/ml IFNγ +10 to 50 μM of each of 69 test compounds (Table S1). Mean ± SEM (n = 3). Atorvastatin and tofacitinib (highlighted) are discussed in the main text.

Figure 4

IFN-inducible PARP9 and DTX3L are required for IFN-induced ADP-ribosylation.A, quantification of nuclear STAT1 phospho-Y701 immunofluorescence signal intensity in A549 cells 24 h after treatment with vehicle control or 100 U/ml IFNγ, with or without 10 μM tofacitinib or 10 μM olaparib, as indicated. Mean ± SEM (n = 6, from three separate experiments), ∗∗∗∗p < 0.0001. B, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells 24 h after treatment with vehicle control or 100 U/ml IFNγ, with or without 10 μM tofacitinib or 10 μM olaparib, as indicated. Mean ± SEM (n = 6–10, from three separate experiments), ∗∗∗∗p < 0.0001. C, quantification of ADP-ribose immunofluorescence signal intensity in WT, PARP9 or DTX3L KO RPE1-hTERT cells 24 h after treatment with vehicle control or 100 U/ml IFNγ. Mean ± SEM (n = 11–17, from four separate experiments), ∗∗∗∗p < 0.0001. D, representative image (top) and quantification (bottom) of immunoblot analyses for STAT1 phospho-Y701 and actin loading control in WT, PARP9 KO, or DTX3L KO RPE1-hTERT cells 24 h after treatment with either vehicle control, 100 U/ml IFN γ, or transfection with 0.1 μg/ml poly(I:C). Mean ± SEM (n = 3). E, relative levels of mRNA for OAS1, IRF1, ISG15, and Mx1 genes determined by RT-qPCR in WT, PARP9 KO, or DTX3L KO RPE1-hTERT cells 24 h after treatment with either vehicle control, 100 U/ml IFNγ, or transfection with 0.1 μg/ml poly(I:C), normalized to respective vehicle-treated WT cells. Mean ± SEM (n = 3).

Figure 5

Nsp3 macrodomain expression has no effect on IFN signaling.A, quantification of nuclear STAT1 phospho-Y701 immunofluorescence signal intensity in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain, 24 h after treatment with vehicle control, 100 U/ml IFNγ, or transfection with 0.1 μg/ml poly(I:C). Mean ± SEM (n = 14, from three separate experiments), ∗∗∗∗p < 0.0001. B, relative levels of mRNA for OAS1, IRF1, ISG15, and Mx1 genes determined by RT-qPCR in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain, 24 h after treatment with vehicle control, 100 U/ml IFNγ, or transfection with 0.1 μg/ml poly(I:C), normalized to respective vehicle-treated empty vector control cells. Mean ± SEM (n = 3). C, representative image of immunoblot analyses for STAT1 phospho-Y701, DTX3L, PARP9, FLAG, and actin loading control in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain, 24 h after treatment with vehicle control, 1000 U/ml IFNα, 1000 U/ml IFNβ, 100 U/ml IFNγ, or transfected with 0.1 μg/ml poly(I:C). Same experiment as Fig. S2A, including parts of the same FLAG panel, which came from the same membrane as phospho-STAT1. Actin loading control from same membrane as PARP9. D, quantification of ADP-ribose immunofluorescence signal intensity in A549 cells transduced with empty vector control (e.v) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain, at indicated timepoints after treatment with 100 U/ml IFNγ. Mean ± SEM (n = 3). E, representative image of immunoblot analyses for STAT1 phospho-Y701, DTX3L, FLAG, and actin loading control in A549 cells transduced either with empty vector control (e.v.) or with a lentiviral construct for constitutive expression of FLAG-tagged macrodomain, at indicated timepoints after treatment with 100 U/ml IFNγ.

Figure 6

The macrodomain reverses PARP9/DTX3L-dependent ADP-ribosylation induced by IFN signaling. Schematic representation of the proposed model. IFN signaling promotes STAT1 phosphorylation by JAK kinases and induces the expression of interferon-stimulated genes (ISGs), including PARP9 and DTX3L. This complex is required for downstream ADP-ribosylation of target proteins, which is counteracted by the viral Nsp3 macrodomain. Neither PARP9/DTX3L nor the Nsp3 macrodomain affects the IFN signaling cascade itself.