Small ubiquitin-related modifier (SUMO) protein conjugation onto target proteins regulates multiple cellular functions, including defence against pathogens, stemness and senescence. SUMO1 peptides are limiting in quantity and are thus mainly conjugated to high-affinity targets. Conjugation of SUMO2/3 paralogues is primarily stress inducible and may initiate target degradation. Here we demonstrate that the expression of SUMO1/2/3 is dramatically enhanced by interferons through an miRNA-based mechanism involving the Lin28/let-7 axis, a master regulator of stemness. Normal haematopoietic progenitors indeed display much higher SUMO contents than their differentiated progeny. Critically, SUMOs contribute to the antiviral effects of interferons against HSV1 or HIV. Promyelocytic leukemia (PML) nuclear bodies are interferon-induced domains, which facilitate sumoylation of a subset of targets. Our findings thus identify an integrated interferon-responsive PML/SUMO pathway that impedes viral replication by enhancing SUMO conjugation and possibly also modifying the repertoire of targets. Interferon-enhanced post-translational modifications may be essential for senescence or stem cell self-renewal, and initiate SUMO-dependent proteolysis.
Small ubiquitin-related modifier (SUMO) proteins can be conjugated onto specific lysine residues of a wide variety of proteins, creating branched peptides1,2. Protein-conjugated SUMO binds SUMO-interacting motifs, inducing conformational changes and stabilizing intra- or intermolecular interactions3. Poly- or multi-sumoylation may also initiate polyubiquitination and proteasome-mediated degradation4,5. At present, sumoylation efficiency is primarily regulated by the target (presence of a canonical sumoylation site or a SUMO-interacting motif) or by E3 enzymes6,7. Most of SUMO1 is conjugated to high-affinity targets, notably to RanGap1, reflecting the fact that SUMO1 abundance is a limiting factor for conjugation. Conversely, SUMO2/3 conjugation is primarily inducible by stress8. Sumoylation is essential for life9 and regulates many biological processes, particularly stress responses10,11. We recently demonstrated a role for PML nuclear bodies (NBs), spherical structures organized by PML protein, in oxidative stress-responsive sumoylation12. There is significant evidence that SUMO conjugation opposes viral or bacterial replication and pathogens inactivate SUMO conjugation through multiple mechanisms13,14,15. Yet, the mechanisms underlying global control of SUMO expression and conjugation remain ill-understood.
Type I or II interferons (IFNs) are cytokines, activated among others by Toll-like receptors (TLRs), involved in response to virus infection, inflammation, immune signalling or stem cell control. Interestingly, multiple proteins involved in IFN signalling or downstream IFN-stimulated proteins are sumoylated16,17. Among IFN/STAT1/3 targets is the RNA binding protein Lin28B, a key developmental regulator implicated in stemness, cancer growth and metabolism18,19. Lin28 is one of the essential human genes contributing to iPSC reprogramming, and Lin28 expression is enriched in various somatic progenitors, as well as in germline and embryonic stem cells19. Mechanistically, Lin28B selectively blocks the biogenesis of let-7 family micro RNAs (miRNAs), whose downstream targets include mitogenic cytokines and oncogenes such as Ras, or Caspase 3.
Here we report that expression of SUMO paralogues is under tight control of the IFN/Lin28/let-7 axis. Such dramatic increases in SUMO abundance induced by IFN treatment or Lin28 expression should profoundly change, not only the abundance of conjugated proteins, but also the repertoire of modified proteins, particularly for SUMO1. Conjugation by SUMOs contributes to the antiviral effects of IFNs and could also be a key player in the control of differentiation and senescence.
IFNs and activated TLR enhance global sumoylation
A 24-h treatment with IFNα initiated a massive increase in SUMO1 and –2/3 conjugated species in different cell lines, as well as in vivo in mice (Fig. 1a–d). The same was observed for IFNα or γ in mouse embryonic fibroblasts (Fig. 1c,d). The levels of both unconjugated (arrows in Fig. 1c,d) and conjugated SUMOs sharply rose. Increase of unconjugated SUMOs after exposure to IFNα was much clearer on ectopic expression of the SUMO protease SENP1 that reverses SUMO conjugation (Fig. 1e). This implies that IFNs primarily increase the expression of SUMO peptides, and not only sumoylation efficacy. We also examined primary mouse or human monocytes stimulated with agonists of TLR, which activate IFN and nuclear factor-κB signalling. Again, IFNα or a TLR9 agonist peptide, but not a control non-TLR9 agonist, sharply enhanced global sumoylation (Fig.1f,g). Increase in SUMO conjugates was often more pronounced for high molecular weight species, which may point to the induction of SUMO chains. Collectively, IFN signalling increases SUMO levels in multiple cell types.
The Lin28 and let-7 axis controls SUMO levels
SUMO1, SUMO2 or SUMO3 mRNA expression was not increased by IFNα, suggesting the involvement of mechanism(s) other than transcriptional control (Supplementary Fig. 1a,b)20. Both IFNs and nuclear factor-κB cascades activate the STAT3/Myc/Lin28B axis, repressing expression of miRNAs from the let-7 family18,21,22,23. Indeed, in HeLa cells, 24-h IFNα treatment inhibited let-7i and let-7g expression by 80 and 60%, respectively, while miR-21 was unaffected (Fig. 2a). In mammals, miRNAs downregulate mRNA stability and protein synthesis through complementary elements in the 3′-untranslated regions (UTRs) of their target mRNAs24. We identified let-7 family complementary sites (LCS) in 3′-UTRs of human SUMO1, SUMO2 and SUMO3 transcripts. SUMO1, SUMO2 and SUMO3 3-′UTRs had nine, seven and two predicted LCS, respectively (Fig. 2b and Supplementary Fig. 1c). Note that the duplexes formed between let-7 family miRNAs and SUMO 3′-UTR LCS primarily contained bulged sequences rather than perfect complementarity (Supplementary Fig. 1c), supporting a silencing mechanism relying on translational downregulation rather than mRNA degradation24. In contrast, we found no LCS on ISG15, glyceraldehyde 3-phosphate dehydrogenase or tubulin transcripts. Similar to human SUMO genes, 3′-UTRs of mouse SUMO1, 2 and 3 transcripts also contained multiple LCSs pointing to evolutionary conservation and physiological importance of this let-7-mediated silencing mechanism (Supplementary Fig. 2a,b).
Inhibition of endogenous let-7i and let-7g expression by specific let-7 antagomirs (HI, hairpin inhibitor), but not by a non-targeting control antagomir, led to global upregulation of sumoylation, mimicking the effects of IFNα (Fig. 2c and Supplementary Fig. 2c) and demonstrating a basal repressive activity of these miRNAs on SUMO expression. Conversely, ectopic overexpression of let-7i and let-7g decreased basal or IFNα-enhanced sumoylation (Fig. 2d and Supplementary Fig. 2c). As expected, expression of other known let-7 targets such as Ras or Caspase 3 were upregulated by let-7 antagomirs or suppressed by let-7 mimics, respectively (Fig. 2e). Essentially similar results were obtained when analysing SUMO1 and SUMO2/3 (Fig. 2c,d). Furthermore, fusion of a complementary DNA encoding green fluorescent protein (GFP) to the non-coding 3′-end of the SUMO1 or SUMO2 genes confirmed the role of the SUMO 3′-UTR in conferring IFNα sensitivity (Fig. 2f), as well as direct let-7 targeting (Fig. 2g). Thus, IFN-enhanced sumoylation involves let-7.
Decrease of SUMO abundance by Lin28 loss or differentiation
Lin28 may indirectly enhance SUMO expression by antagonizing let-7 expression. Indeed, directly inactivating Lin28 expression by small interfering RNA (siRNA) led to a sharp decrease in both unconjugated and conjugated SUMO1 and SUMO2/3 (Fig. 3a). Thus, the Lin28/let-7 axis controls basal and IFN-enhanced SUMO1/2/3 expression.
Lin28 proteins have recently emerged as key factors defining stemness in various lineages25,26,27, questioning the existence of a gradient of SUMO expression during normal differentiation. To address this issue, we isolated mouse haematopoietic progenitors and compared their sumoylation profile with that of total marrow primarily containing their differentiated progeny. In haematopoietic progenitors, as defined by the absence of expression of any lineage markers (Lin−), we observed a dramatic enrichment of unconjugated SUMO1 and SUMO2/3 (Fig. 3b and Supplementary Fig. 3a). SUMO1, but not SUMO2/3, conjugates were also clearly upregulated, suggesting that in these primary cells, SUMO2/3 may require a stress signal to undergo conjugation.
Lin28 upregulation was also implicated in cellular transformation. For example, Lin28 levels decrease rapidly during all-trans retinoic acid (ATRA)-induced differentiation of mouse embryonic carcinoma cells28, and during ATRA-induced differentiation of human promyelocytic leukemia NB4 cells, let-7 miRNA levels increase29. As predicted, NB4 differentiation led to a dramatic decrease in Lin28B levels and a sharp decline in SUMO1 and SUMO2/3 conjugates (Fig. 3c and Supplementary Fig. 3b). Collectively, these results demonstrate that decreased Lin28 levels correlate with impaired sumoylation, possibly contributing to Lin28-mediated regulation of stemness and differentiation.
IFN-induced SUMOs are required for IFN antiviral effects
SUMOs oppose the replication of viruses or bacteria, raising the possibility that they mediate, at least in part, the effects of IFN on pathogens13,14. Several viral or bacterial proteins downregulate global sumoylation. For example, the Herpes virus ICP0 protein targets sumoylated proteins to the proteasome30 (Fig. 4a). Interestingly, only ICP0-defective HSV1 is susceptible to IFNα31 (Fig. 4b and Supplementary Fig 4a). To formally demonstrate the role of IFNα-induced SUMOs in the antiviral effects of this cytokine, we compared viral replication in U2OS cells transfected with siRNAs against SUMO1/2/3 and then treated or not with IFNα, before infection with wild-type or ICP0-defective HSV1. Downregulation of SUMOs favoured (<3-fold) HSV1 replication, confirming their role as antiviral peptides (Fig. 4b and Supplementary Fig. 4a,b). However, a striking effect of SUMO downregulation was observed on IFNα-treated cells. Although IFN inhibited replication of the ICP0-defective mutant virus (by 59-fold, Fig. 4b), this antiviral effect was blunted on silencing of SUMOs (12-fold reduction). Thus, SUMOs are critical downstream mediators of IFN antiviral effect.
IFN-inducible PML protein has intrinsic antiviral activity32,33 and promotes sumoylation of specific substrates recruited on PML NBs12. This suggested that PML might redirect IFN-induced SUMOs towards PML NB partners, such as SP100, rather than non-NB-partner proteins such as RanGAP1. Indeed, in the absence of PML, RanGAP1 sumoylation was considerably more enhanced by IFN (Supplementary Fig. 4c), while efficient IFN-induced SP100 sumoylation is known to require PML12. IFN antiviral effect against HSV1 was impeded by PML silencing (11-fold reduction, Fig. 4b and Supplementary Fig. 4b), in keeping with previous reports31,34. Importantly, PML and SUMO inactivation did not synergize, strongly suggesting that they act on the same pathway. Thus, SUMOs and PML cooperate to enforce hypersumoylation of targets that are essential for IFNα-initiated anti-HSV1 effects.
We finally investigated the role of SUMO induction on the replication of two retroviruses, HIV and B-Murine Leukemia Virus. Again, IFNα suppressed infectivity of both viruses by approximately fourfold, an effect which was significantly impeded (>50%) on inactivation of SUMOs (Fig. 4c and Supplementary Fig. 4d). Thus, IFN-triggered hypersumoylation directly contributes to the protection against a variety of viruses.
In general, post-translational modifications are primarily regulated by interactions between protein targets and their modifying enzymes, rather than by the availability of the molecule to be conjugated. Here we unravel major changes in the abundance of SUMOs in IFNα/γ response, and link them to Lin28/let-7 cascade (Fig. 4d). Another less-studied IFN-responsive ubiquitin-like peptide, ISG15, was recently implicated in control of mycobacterial infection35,36,37. Thus, apart from target gene induction by STATs and IRFs, which play non-overlapping roles in antiviral activities38, activation of ubiquitin-like post-translational modification may be key, global and previously underappreciated, effectors underlying IFN responses.
In normal cells, SUMO1 is limiting and almost entirely conjugated to RanGAP1 (ref. 1). On IFN treatment or in progenitors, unconjugated SUMO1 becomes abundant (Figs 1c–f and 3b), making it available for the conjugation of low-affinity targets. Thus, our observations predict a profound change in the repertoire of SUMO1-conjugated proteins on IFN exposure. PML, another IFN-induced protein39, promotes sumoylation by favouring interactions between UBC9, the universal SUMO E2 ligase and specific partners within PML NBs12. Several of these partners are themselves transcriptionally controlled by IFN, notably SP100 (ref. 40). Often in their sumoylated forms, some partners have been implicated in IFN effects on senescence (DAXX, p53, SP100) or antiviral defence (SP100)41,42,43,44,45,46. These unexpected findings therefore point to the existence of a highly integrated regulatory loop wherein the substrates, their modifiers (SUMOs) and the catalyser (PML NBs) are all dramatically upregulated during IFN response, promoting the rapid formation of the active, SUMO-conjugated form. This integrated IFN-driven pathway is targeted by multiple viral or bacterial proteins, which abrogate de novo sumoylation, promote desumoylation, degrade SUMO conjugates or de-organize PML NBs15,47.
The evolutionary conservation of the miRNA-based regulation points to the importance of the regulatory loop identified here. Apart from virus infection, there is circumstantial evidence that IFN control over SUMO or PML abundance may be also important for other processes. Early progenitors are very SUMO rich, consistent with studies that have linked SUMOs to stem cell maintenance48,49,50. PML was also implicated in regulation of stem cell fate51, notably through TR2 sumoylation, which directly controls expression of Oct-4 (refs 52, 53), or through p53 activation54. Thus, IFN-initiated, PML-enforced hypersumoylation may control stem cell self-renewal and/or proliferation42,55. Lin28 controls metabolism19,56, as recently shown for PML57,58. Future studies should clarify any implication of SUMOs in these processes. Finally, PML or SUMOs were repeatedly associated with the control of IFN signalling through STAT1 or IRF3/7 conjugation16,17,59,60, suggesting the existence of a feed-back loop.
As hypersumoylation can promote polyubiquitination4, our findings also have important implications for regulation of proteolysis. IFNs induce a transient global polyubiquitination61, which could result from the hypersumoylation unravelled here. These findings have important therapeutic relevance, as IFNα promotes PML- and SUMO-dependent degradation of toxic proteins in the brain62,63 or of the HTLV-I leukemia oncoprotein Tax64, with unambiguous clinical benefit62,65,66. Thus, similar to therapies based on manipulation of ubiquitin-dependent degradation67,68, this could pave the way to IFN-triggered, SUMO-based therapies promoting the proteolytic clearance of undesirable proteins68.
Bioinformatics and modulation of microRNA expression
Binding sites for let-7 miRNA family members on human and mouse SUMO1, 2 and 3 transcripts and predicted duplex formations were determined using miRanda application69. miRIDIAN mimic hsa-let-7g or hsa-let-7i, miRIDIAN HI hsa-let-7g or hsa-let-7i or scrambled control oligos were from Dharmacon. Expression levels of mature let-7g, let-7i and miR-21 were determined by quantitative PCR using miRCURY LNA universal cDNA synthesis kit, miRCURY LNA SYBR® Green master mix and hsa-let-7g-, hsa-let-7i-, hsa-let-7g*-, hsa-let-7i*- and hsa-miR-21-specific primers (Exiqon). U6 small nuclear RNA was used as an internal control. Expression levels of murine SUMO1, SUMO2, SUMO3 and PML transcripts were determined by quantitative PCR using Taqman probe-based gene expression analysis (Applied Biosystems), where a glyceraldehyde 3-phosphate dehydrogenase probe was used as an internal control.
Expression vector for Senp1 was a gift from E.Y. Yeh and expression vector for ICP0 from B. Roizman. 3′-UTRs of SUMO1 and SUMO2 genes were amplified by PCR from HeLa genomic DNA (SUMO1 3′-UTR forward: 5′-TTT ATT TCT AGA AAT AGT TCT TTT GTA ATG TGG-3′, SUMO1 3′-UTR reverse: 5′-CAT GCC TCT AGA TTC AAC ATG ATT AGG TAA CTG-3′; SUMO2 3′-UTR forward: 5′-GGT GTC TCT AGA AAA GGG AAC CTG CTT CTT TAC-3′, SUMO2 3′-UTR reverse: 5′-ACT ACT TCT AGA TCA TTT TAA ACA AGA AGT TTA TTT AAA C-3′) and inserted into the XbaI site of pEGFP-N1 (Clontech). As IFN downregulates expression from the cytomegalovirus promoter, expression levels of SUMO 3′-UTR–GFP fusions were normalized to the unfused GFP control. Basal expression levels of GFP fusions are lower than that of the unfused GFP control in accordance with basal cellular suppressive effects of endogenous let-7 miRNAs. For simplicity, basal expression levels of all GFP constructs (fused or unfused) were normalized to 1 in Fig. 2f,g.
Mouse monoclonal anti-SUMO1 (21C7) and rabbit polyclonal anti-SUMO2/3 antibodies were from Invitrogen. All primary antibodies were used for western blot analysis at 1/1,000 in TBS-0.1% Tween 20. Mouse monoclonal anti-GFP antibody was purchased from Roche (clones 7.1 and 13.1) and rabbit polyclonal anti-Lin28B antibodies were from Cell Signaling (4196S) and Abcam (ab71415). Goat polyclonal anti-human Caspase 3 (L-18) and mouse monoclonal anti-vinculin (7F9) antibodies were from Santa Cruz, rabbit polyclonal anti-human N-Ras (N5385-01D) antibody was from US Biological. Chicken anti-human PML and rabbit polyclonal anti-SP100 antibodies are homemade antibodies. Mouse monoclonal anti-mouse PML was kindly provided by Scott W. Lowe. All antibodies were revealed by horseradish peroxidase-conjugated secondary antibodies (1:20,000, from Jackson). SUMO conjugates were separated on pre-casted 4–12% gradient gels (Invitrogen).
Cell culture and treatments
HeLa cells were cultured in DMEM medium supplemented with 10% FCS. siRNAs against Lin28A and B were from Qiagen (Hs Lin28_5, Hs Lin28_6, Hs Lin28B_1 and Hs Lin28B_2). siRNAs against SUMO1, 2 or 3 were from Qiagen4. Let-7 HI or mimics were transfected using DharmaFECT-1 Transfection Reagent (Dharmacon) according to the manufacturer’s guidelines, at 25 nM where not indicated. Cells were lysed 72 h post transfection for protein analysis. IFNα or γ (human IFNα, Roche; mouse IFNα and γ, Hycult Biotech) were used at 1,000 and 500 IU ml−1, respectively, for the indicated period of time. ATRA exposure on NB4 cells, cultured at 2 × 105 ml−1 in RPMI 10% fetal bovine serum, was performed at 10−6 M for 48 h. Differentiation of cells was analysed by MGG (May–Grünwald Giemsa) staining. Lin− cells were obtained from fresh mouse bone marrow by depletion using rat anti-mouse CD4, CD8, MacI, GrI, B220 antibody cocktail on anti-rat IgG Dynabeads (Life Technologies). FACS on bone marrow and lin− cells was then performed with the same cocktail to ensure enrichment. Primary mouse monocytes were isolated on depletion of other lineages from blood using anti-B220 (CD45R), anti-GR1 (Ly6G, Ly6C), anti-CD5 (Ly1) and anti-CD8a (Ly2) antibodies (Miltenyi Biotech), and kept in culture in RPMI medium supplemented with 10% FCS, interleukin-3, interleukin-6 and Stem Cell Factor for 24 h. Primary human monocytes were isolated from human peripheral blood mononuclear cells using the Monocyte Isolation Kit II (Miltenyi Biotech), and the purity of enriched monocytes was evaluated by flow cytometry, according to manufacturer’s instructions. Human and mouse TLR9 agonists (ODN2216 and ODN1585) and the non-agonist controls were from InvivoGen and used at 5 × 10−6 M final concentration for 24 h.
In vivo treatments
Mouse handling was performed in accordance with established institutional guidance and approved protocols from the Comité Régional d’Ethique Expérimentation Animale n°4, which enforces the EU 86/609 directive. Female FVB-Nico mice of 8 weeks were treated with 106 IU murine IFNα for 1 day, after which they were killed by cervical dislocation, organs were collected and protein extracts analysed by western blotting.
Viral infections and statistical analyses
Wild-type and ICP0-deleted R7910 HSV1 were obtained from B. Roizman (University of Chicago). U2OS cells, which were chosen for efficient replication of ICP0-defective virus31, were cultured in standard medium, transfected twice with siRNA against SUMO1/2/3, PML or control siRNA, and treated then with IFNα (1,000 IU ml−1) for 24 h before infection. One day post infection, cells were lysed and virus titres determined by plaque assays. Comparison of virus titres was performed using two-tailed Student’s t-test, assuming unequal variances.
HIV and B-murine leukemia virus viral stocks were produced by transfection of 293T cells using a standard calcium–phosphate precipitation technique. Briefly, cells were transfected with plasmids encoding the HIV or MLV packaging DNA, the genomic transfer vector encoding GFPor RFP (red fluorescent protein) and an expression vector for the glycoprotein G of vesicular stomatitis virus. Supernatants were collected 40 h post transfection, clarified by low-speed centrifugation and filtered through 0.45-μm pore size filters. HIV viral particles were concentrated by ultracentrifugation (24,000 r.p.m., 90 min, 4 °C) using a SW32 rotor (Beckman) on a 20% sucrose cushion. All viral stocks were titrated by infection of 293T cells. After 48 h, the percentage of GFP- or RFP-expressing cells was measured by flow cytometry on a FACSCalibur flow cytometer with CellQuest software (BD Biosciences). HeLa cells (20,000 cells per well in 24-well plates) were then transfected with siRNA against SUMO1/2/3 or control non-targeting siRNA and treated with IFNα (1,000 IU ml−1) for 24 h before infection with glycoprotein G of vesicular stomatitis virus-pseudotyped HIV or MLV viral particles. Infectivity was scored by measuring the percentage of cells expressing the GFP or RFP reporter gene by flow cytometry 24 h post infection.
How to cite this article: Sahin, U. et al. Interferon controls SUMO availability via the Lin28 and let-7 axis to impede virus replication. Nat. Commun. 5:4187 doi: 10.1038/ncomms5187 (2014).
The laboratory is supported by INSERM, CNRS, Université Paris-Diderot, Institut Universitaire de France, Ligue Contre le Cancer, Institut National du Cancer, ANR (PACRI and SLI projects), Association pour la Recherche contre le Cancer (Griffuel Award to HdT), Canceropôle Ile de France and the European Research Council (STEMAPL advanced grant to HdT). U.S. was supported by a fellowship of the Fondation pour la Recherche Medicale. The data were presented at the EMBO meeting ‘Ubiquitin and ubiquitin-like proteins: From structure to function’ in October 2013. We thank B. Roizman for Herpes viruses and S. Lowe for the anti-pml monoclonal antibody. We thank all members of the laboratory for help and discussions, and A. Saïb, K. Rice and A. Bazarbachi for critical review of the manuscript.
Supplementary Figures 1-5
About this article
Nature Communications (2016)