Stat1 and SUMO modification

Characterization of the ability of type I interferons (IFNs) (eg, IFN-α) to rapidly activate genes led to the identification of ISGF-3, a transcription factor consisting of Stat1, Stat2, and an IFN regulatory factor-9 (IRF-9) DNA binding protein.¹  Subsequently, IFN-γ was shown to induce genes through Stat1 homodimers.²  To date, 7 signal transducers and activators of transcription (STATs) have been identified in vertebrates, all of which are activated by phosphorylation on a single tyrosine (Y701 in Stat1; reviewed in Levy and Darnell³  and Kisseleva et al⁴ ). Activation drives STAT dimerization by directing a stable and specific association between the phosphotyrosine of one STAT and the src homology 2 (SH2) domain of a partner STAT.⁵  Residues located at positions +1, +3, +5, +6, and +7 carboxy terminal to this phosphotyrosine (ie, amino acids 702, 704, 706, and 707 for Stat1) determine the specificity of this interaction.⁶  Dimerized STATs translocate to the nucleus, where they bind to members of the gamma-activated site (GAS) family of enhancers, culminating in the induction genes.^3,4 

The regulation of STAT signal decay has also been an area of active investigation. Four major classes of counterregulatory molecules have been identified, including phosphatases,^3,4,7  nuclear “transportases,”^8-10  covalent modifiers,^4,11,12  and specific STAT counterregulatory proteins (eg, suppressor of cytokine signaling [SOCS] and protein inhibitor of activated STATs [PIAS] proteins^13,14 ). Studies on SOCS-1 have provided significant evidence for a critical role in down-regulating IFNγ-Stat1–dependent signals, but studies on PIAS proteins have yielded less direct mechanistic insight into Stat1 regulation.^14-16  More recent studies have determined that PIAS proteins are small ubiquitin-related modifier (SUMO) E3 ligases, raising the possibility that STAT activity is regulated through SUMO modification.^17-19 

SUMOs are approximately 100–amino acid peptides which, like ubiquitin, become covalently attached to cellular target proteins (reviewed in Kim et al,¹⁷  Melchior et al,¹⁸  and Müller et al¹⁹ ). However, in contrast to ubiquitin, SUMO modifications do not target proteins for degradation, but rather promote protein–protein interactions and direct subcellular localization, and/or serve to antagonize ubiquitin-dependent degradation. SUMO conjugation entails the formation of a reversible isopeptide bond between the C-terminus of the SUMO peptide (SUMO-1, SUMO-2, or SUMO-3) and the ϵ amino group of the lysine found in the consensus sequence ψKxE (ψ indicates hydrophobic residue, and x indicates any residue; Table 1). Analogous to ubiquitin, SUMO conjugation is mediated by an ATP-dependent E1-activating complex (ie, Aos1 + Uba2), an E2 ligation complex (ie, Ubc9) and an E3 conjugation complex. The relative specificity exhibited by Ubc9 for some SUMO substrates is likely to account for E3-independent SUMO conjugation observed in vitro.^20,21  Finally, isopeptidases from the SUSP/SENP family assure that SUMO modification is reversible.^18,22 

Sequence analysis revealed 2 potential SUMO modification sites, ¹⁰⁹LKEE¹¹² and ⁷⁰²IKTE⁷⁰⁵, in Stat1, but not in other STATs (Table 1). In vitro SUMO conjugation studies determined that Stat1 is SUMO modified at lysine 703, but not lysine 110. A subsequent functional analysis of 2 SUMO-ylation–resistant Stat1 mutants, Stat1^K703R and Stat1^E705A, revealed 2 distinct phenotypes. Stat1^K703R exhibited enhanced DNA binding, prolonged nuclear retention, and modest changes in the biologic response to IFN-γ, as recently reported.²³  In contrast, Stat1^E705A exhibited wild-type DNA binding, nuclear retention, and biologic response to IFN-γ. These observations suggest that lysine 703, located at the critical interface between Stat1 homodimers,⁶  plays a fortuitous and structurally important role in Stat1 DNA binding activity, and that Stat1 activity is not likely to be significantly regulated by SUMO conjugation in vivo.

HEK-293T, Vero, and L929 cells were from ATCC (Manassas, VA); and Stat1^–/– mouse embryonic fibroblasts (MEFs) and macrophages were harvested from Stat1^–/– mice (generously provided by D. Levy, New York University, NY). Cells were cultured in Dulbecco modified Eagle medium, supplemented with 10% fetal bovine serum from GIBCO Laboratories (Grand Island, NY). Bone marrow (from mouse femurs)–derived macrophages were cultured in 20% L929 cell–conditioned media for 5 to 10 days.²⁴  After 16 hours in culture, adherent bone marrow cells were infected (3 times) with retrovirus freshly prepared from HEK-293T transfectants (below). Stat1^–/– MEFs were infected twice with pMIG retroviruses encoding Stat1, Stat1^K703R, and Stat1^E705A in the presence of polybrene (8 μg/mL; Sigma, St Louis, MO), as previously reported.²⁵  High-titer viral supernatants were prepared through transient transfection in HEK-293T cells by calcium phosphate precipitation.⁹  Retroviral infection efficiency, as determined by fluorescence-activated cell-sorter (FACS) (green fluorescent protein–positive [GFP⁺] cells; FACS-Calibur; BD Biosciences, San Jose, CA), varied between 90% to 95% in MEFs and 25% to 35% in bone marrow macrophages (BMMs).

For viral response assays, MEFs were infected with vesicular stomatitis virus (VSV, Indiana strain; gift from R. Pine, Public Health Research Institute, Newark, NJ) prepared in Vero cells. Viral yield was measured 24 hours after infection by titering on Vero cells overlayed with 1.5% methyl cellulose for 36 hours.²⁶  Expression of the reporter GFP gene and major histocompatibility complex (MHC) II (after staining with anti-mouse I-A^b; BD Pharmingen, San Diego, CA) was evaluated by flow cytometry (FACS-Calibur).^9,26  Nitric oxide (NO) production was evaluated at day 10 of culture and 72 hours after IFN-γ (50 U/mL) and/or LPS (2 μg/mL) stimulation, as previously described.²⁷ 

Recombinant SUMO-1, E1, and E2 were expressed, purified, and assayed as previously described.²⁰  The IR1-M-IR2 domains of Nup358 (aa 2596-2836) and full-length PIAS1 (aa 1-651) were cloned and expressed as Smt3 fusion proteins, cleaved by Ulp1, and purified by anion exchange and gel filtration chromatography.²⁸  To assay for Stat1 SUMO-ylation, purified recombinant Stat1 (0.5 μM) was incubated with SUMO-1 (2 μM), E1 (Uba2/Aos1; 0.3 μM), E2 (Ubc9; 0.3 μM), and E3 (PIAS1 or Nup358 at 0.3 μM) in a buffer with 5 mM MgCl₂, 20 mM HEPES (pH 7.5), 1 mM DTT, 2 mM ATP, and 0.5 U pyrophosphatase (Sigma) for 1 hour at 37°C. To assay for peptide SUMO-ylation, p53 (residues 323-393), p53^K386M (residues 320-393) Stat1/phospho-Stat1 (residues 697-711), or Stat3 (residues 701-714) peptides (at 500 μM) were incubated with E1 and either wild-type or mutant Ubc9 for 5 or 24 hours at 37°C, as previously described.²⁹  Samples were fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and stained with Coomassie Blue or Sypro (4-12–hour staining, > 1 hour destaining; Bio-Rad, Hercules, CA). Sypro-stained gels were imaged by UV illumination and the band intensity was quantified (Bio-Rad UV system; Quantity One Software; Bio-Rad).

Overexpression assays entailed transfecting protamine complementary DNA (pcDNA; Invitrogen, Carlsbad, CA) expression plasmids encoding hemagglutinin (HA)–SUMO-1, His–SUMO-1, Ubc9, and/or PIAS-y (gifts from S. Goff, Columbia University, New York, NY) into HEK-293T cells by calcium phosphate precipitation or MEFs with LipofectAmine Plus regent (Invitrogen). For oligonucleotide “pull-down” assays 20 μg GAS oligonucleotide (Table S1, available on the Blood website; see the “Supplemental Table” link at the top of the online article) coupled to 75 μL biotinylated agarose beads (Sigma) was incubated with 400 μL whole-cell extracts (WCEs) for 2 hours at 4°C, after blocking the beads with 1% bovine serum albumin (BSA) for 1 hour at 4°C, as previously reported.³⁰  For immunoprecipitation, 400 μL WCEs was incubated with primary antibody for 2 to 16 hours followed by protein A agarose (Sigma), and then collected, as previously reported.^9,26  WCE or precipitates were fractionated by SDS-PAGE and then evaluated by immunoblotting with the appropriate antibody.⁹  For nickel pulldown, 800 μL WCEs was incubated with 100 μL ProBond nickel beads, washed, and eluted as per the manufacturer's instructions (Invitrogen). For electrophoretic mobility shift assay (EMSA), extracts were incubated with a GAS probe and fractionated by native PAGE as previously described.^9,26  Antibodies were directed against Stat1,¹  phosphotyrosine Stat1 (Cell Signaling Technology, Beverly, MA), MHC II (BD Pharmingen), HA (Covance, Berkeley, CA), and β-actin (Sigma).

pClEco and pMIG were provided by J. Luban (Columbia University).²⁵  Murine Stat1, Stat1^K703R, and Stat1^E705A were cloned into the XhoI site in pMIG. Stat1^K703R and Stat1^E705A were prepared by site directed mutagenesis (Quickchange kit; Stratagene, La Jolla, CA) and confirmed by sequencing (Table S1 for primer sequences).

Cells were cultured on sterile cover slips until 20% to 25% confluent, fixed in formaldehyde, and stained as previously reported⁹  with Stat1-specific antibodies (1:250 fold dilution) and a Cy3-conjugated secondary antibody (1:500; Jackson ImmunoResearch Laboratories, West Grove, PA). Slides were examined under a Nikon Eclipse TE300 microscope (Nikon, Melville, NY) after excitation at 550 nm (Cy3) and excitation at 495 nm (GFP).

Total RNA was prepared from MEFs with Trizol (Invitrogen) extraction. RNA (4 μg) was treated with RQ1 DNAse (Promega, Madison, WI) and then reverse-transcribed with SuperScript II (Invitrogen). The cDNA was quantitatively amplified in an ABI Prism 7700 polymerase chain reaction (PCR) system with a SYBR green master mix (Applied Biosystems, Foster City, CA) as previously described.²⁴  Gene expression was normalized to a β-actin control.

MEFs were seeded in 24-well plates (7.5 × 10⁵/well) and transfected with 400 ng of an IRF-driven luciferase reporter (ie, B2WT3³¹ ) and 40 ng pRL-tk renilla in LipofectAmine Plus (Invitrogen). Later (24 hours), transfectants were stimulated with IFN-γ (5 U/mL for 6 hours), resuspended in a passive lysis buffer, and evaluated in a luminometer (TD 20/20; Turner Systems, Sunnyvale, CA), as previously reported.³¹  Experiments were done in triplicate and luciferase activity was normalized to renilla activity.

Recent studies identifying PIAS proteins as SUMO E3 ligases suggested that STATs may also be SUMO modified.^17-19  Analysis of all mammalian STAT sequences identified 2 potential SUMO modification sites: Stat1 residues ¹⁰⁹LKEE¹¹² and ⁷⁰²IKTE⁷⁰⁵ (Table 1). Efforts to exploit SUMO-1–specific antibodies to detect endogenous SUMO-modified Stat1 were unsuccessful. To improve the chances of detecting this product, Ubc9 and HA–SUMO-1 were overexpressed in HEK-293T cells. Under these conditions SUMO-Stat1 conjugates were readily recovered in Stat1 immunoprecipitates (Figure 1A) as previously reported.^32,33  Coexpression of PIAS-y, however, impeded recovery of SUMO-Stat1 conjugates, most likely because of reduced levels of SUMO-1 and Ubc9 expressed in triple transfectants.

One potential SUMO modification site, lysine 703 (K703), lies adjacent to the tyrosine that is phosphorylated during Stat1 activation, (ie, Y701). It was therefore important to determine whether SUMO conjugation affected Stat1 phosphorylation and subsequent DNA binding activity. As anticipated, phosphorylated Stat1 was readily recovered, by either immunoprecipitation or oligonucleotide precipitation, from prepared extracts of IFN-γ–stimulated HEK-293T cells overexpressing Ubc9 and HA–SUMO-1 (Figure 1B, left panel). Although slower migrating SUMO-Stat1 conjugates were readily recovered in the Stat1 immunoprecipitates, they were not detected in the oligonucleotide precipitates, suggesting that SUMO-modified Stat1 will not bind DNA. Likewise, phosphorylated Stat1 was readily recovered in Stat1 immunoprecipitates prepared from IFN-γ–stimulated HEK-293T cells overexpressing Ubc9 and His–SUMO-1. Yet, even though Stat1 was readily collected by nickel–His-SUMO-1 pull-down and phospho-Stat1 was abundant, no phospho-Stat1 was recovered in the nickel pull-downs (Figure 1B, right panel). These observations suggest that Stat1 modification by tyrosine phosphorylation and SUMO-ylation are mutually exclusive.

To develop more direct evidence for Stat1 SUMO modification and to map the SUMO-ylation site, we turned to an effective in vitro conjugation assay that overcomes limitations imposed by SUMO deconjugating enzymes.^20,22,34  Purified recombinant Stat1 was incubated with purified active preparations of human recombinant SUMO-1, E1 (Uba2/Aos1), E2 (Ubc9), and E3 (Nup358 or PIAS1).^20,21,35  After 1 hour, the products were fractionated by SDS-PAGE and immunoblotted with a Stat1-specific antibody. As shown in Figure 2, the anticipated approximately 105-kDa SUMO-Stat1 conjugate was readily formed in an ATP-dependent, but not E3-dependent manner (Figure 2A). Analogous results were obtained with a purified p53 control (data not shown). The SUMO-Stat1 conjugates (approximately 3%-4% of the total Stat1) and the remaining nonreacted (ie, native) Stat1 were then excised, digested with trypsin, and evaluated by mass spectrometry. The specific loss of peptides spanning K703, but not K110 indicated that Stat1 had been SUMO modified at K703 (data not shown). To confirm that K703 was the SUMO-ylation site, a Stat1 peptide spanning residues 697-711 (ie, KGTGYIKTELISVS) was evaluated in an in vitro SUMO conjugation system where a defective E2 (ie, Ubc9^C93S) served as a negative control.²⁹  Several additional peptides were evaluated. This included the analogous peptide from Stat3 (ie, AAPYLKTKFICVT) and a Stat1 peptide in which K703 was mutated to arginine (ie, KGTGYIRTELISVS). To explore whether Stat1 Y701 phosphorylation precludes SUMO modification, as suggested by HEK-293T overexpression studies (Figure 1), a Stat1 phosphopeptide (ie, phospho-Y701) was also evaluated. Again, p53 peptides spanning a well-characterized SUMO-ylation site served as important controls.²⁰  After 5 and 24 hours of SUMO conjugation, only the wild-type Stat1 peptide was SUMO-ylated as effectively as the control wild-type p53 peptide (Figure 2B). More detailed kinetic studies provided further evidence that wild-type Stat1 was an effective SUMO substrate, whereas Stat3, phospho-Stat1, and the mutant peptides were poor substrates (Figure 2C). These data confirmed K703 as the SUMO modification site, but only when the adjacent Y701 was not phosphorylated. The failure to SUMO-ylate the Stat3 peptide was consistent with our inability to modify purified preparations of recombinant Stat3⁹  (data not shown), reflecting a critical divergence in the corresponding potential Stat3 SUMO conjugation site (Table 1).

To explore the potential role of Stat1 SUMO modification in vivo, a SUMO-ylation–resistant Stat1^K703R mutant was generated. Retroviral vectors directed expression of Stat1^K703R and wild-type Stat1 at physiologic levels in Stat1^–/– MEFs³⁶  in more than 95% of infected cells (data not shown; Figure 3A-B). Upon brief (ie, 0.5 hours) stimulation with IFN-γ, both Stat1 and Stat1^K703R were rapidly activated (data not shown; Figure 3B). However, when these extracts were evaluated by EMSA, the DNA binding activity of Stat1^K703R was significantly more prolonged than that of wild-type Stat1 (data not shown; Figure 3C). To determine whether Stat1^K703R exhibited an enhanced affinity for GAS elements, DNA dissociation studies (ie, “off-rate”) were performed. Stat1^K703R displayed a 4-fold slower off-rate than wild-type Stat1 (ie, t_1/2 > 120 vs 30 minutes; Figure 4A-B). Likewise, Stat1^K703R exhibited an increased relative GAS binding activity (approximately 4-fold) compared with wild-type Stat1 (Figure 4C-D). As anticipated, no differences were observed in ISGF3-ISRE DNA binding activity in IFN-α–stimulated extracts, where IRF-9 mediates DNA binding (data not shown). In sum, these studies demonstrate that mutation of lysine 703 to arginine significantly enhanced Stat1 GAS binding activity in response to IFN-γ stimulation, recently also reported in human U3A cells.²³ 

To develop additional evidence that Stat1 was SUMO modified at K703 in vivo, a second critical amino acid in the Stat1 SUMO consensus modification site (ie, E705), which is also not involved in dimerization, was mutated to alanine.^6,17,18  Again, retroviral vectors directed a physiologic expression of Stat1^K703R, Stat1^E705A, and wild-type Stat1 in more than 95% of Stat1^–/– MEFs (Figure 3A, top panel). To confirm that both mutants were defective in SUMO modification, HA–SUMO-1 and Ubc9 were overexpressed in these MEFs. Despite modest transfection efficiency in these MEFs (Figure 3A, bottom panel), the studies clearly demonstrate the formation of the approximately 105-kDa SUMO-Stat1 conjugate in cells expressing wild-type Stat1, but not in those cells expressing the SUMO-ylation–defective mutants Stat1^K703R and Stat1^E705A (Figure 3A, top panel).

Next, the IFN-γ–dependent activation of Stat1^E705A was evaluated. Stat1^E705A exhibited a pattern of activation (ie, tyrosine phosphorylation) and DNA binding that was identical to that of wild-type Stat1, with continuous (ie, up to 12 hours) IFN-γ stimulation (Figure 3B-C). Likewise, Stat1^K703R exhibited an essentially normal pattern of tyrosine phosphorylation, notable for a slight delay at early time points and more robust activity at later time points (Figure 3B).

However, these modest changes in IFN-γ–dependent Stat1^K703R tyrosine phosphorylation correlated with a striking increase in DNA binding activity (Figure 3C). Thus, SUMO-ylation–defective Stat1^E705A exhibits an activation profile that is distinct from Stat1^K703R, but similar to wild-type Stat1.

The rapid and transient translocation of Stat1 to the nucleus is another characteristic feature of Stat1 signaling activity. Since several studies have implicated SUMO modification in the regulation of nuclear trafficking,^34,37,38  a set of Stat1 immunolocalization studies were carried out. Analogous to the pattern observed with wild-type Stat1, both Stat1^K703R and Stat1^E705A exhibited a predominately cytoplasmic distribution in unstimulated cells and robust nuclear accumulation with IFN-γ stimulation (ie, 0.5 hours; Figure 5). After stimulation (6 hours), both Stat1 and Stat1^E705A were fully re-exported back to the cytoplasm, rendering those cells ready for another round of stimulation. In contrast, Stat1^K703R continued to exhibit a strong nuclear retention, consistent with its enhanced DNA binding activity.¹⁰  These studies demonstrate that the 2 SUMO-ylation–defective Stat1 mutants exhibit remarkably divergent phenotypes.

Next, several studies were undertaken to explore the potential role of SUMO-ylation in regulating biologic responses directed by Stat1. In the first set of studies, the ability of Stat1^–/– MEFs expressing wild-type Stat1, Stat1^K703R, or Stat1^E705A to direct the IFN-γ–dependent induction of several target genes was explored (Figure 6). Wild-type Stat1 and Stat1^E705A promoted similar normal patterns of IRF1, TAP1, GBP1, and Mig expression.^24,39  However, cells expressing Stat1^K703R exhibited a delayed expression pattern, with significant reduction of target gene expression at 2 and 6 hours after stimulation, followed by normalization of expression by 12 hours (Figure 6A). Changes in the expression of a GAS-driven reporter gene, which tends to measure a more averaged transcriptional response, did not reveal any statistically significant differences between wild-type, Stat1^E705A, or Stat1^K703R (Figure 6B; Figure 6C illustrates that all Stat1s were expressed at equivalent levels). There was, however, a consistent trend toward diminished reporter expression in the Stat1^K703R cells. Thus, Stat1^E705A is functionally indistinguishable from wild-type Stat1, but the transcriptional kinetics in Stat1^K703R cells are notable for a reduced initial expression of target genes in the setting of a mutant with enhanced DNA binding activity.

IFNs stimulate a potent and physiologically important antiviral response, which represents an integrated response of numerous target genes.^4,26  To determine whether SUMO-ylation–defective Stat1 mutants direct an abnormal antiviral response to IFN-γ, Stat1^–/– MEFs expressing either wild-type Stat1, Stat1^K703R, or Stat1^E705A were infected with VSV. In the first study, cells were pretreated with increasing doses of IFN-γ (0.5-50 U/mL) prior to infection with a fixed multiplicity of infection (MOI = 0.5; Figure 7A, top panel). As anticipated, Stat1^–/– cells were not protected by IFN-γ pretreatment,^26,36  but Stat1^E705A and wild-type Stat1 directed the same robust antiviral response to IFN-γ (at 5 and 50 U/mL). Stat1^K703R demonstrated modestly enhanced antiviral activity, but only at low IFN-γ doses. Similarly, Stat1^K703R directed a modestly enhanced antiviral response to IFN-γ (5 U/mL) when the VSV MOI was varied from 0.001 to 1.0 (Figure 7A, bottom panel). Thus, SUMO-ylation–defective Stat1^E705A directs a wild-type antiviral response to IFN-γ. In contrast, Stat1^K703R, with its enhanced DNA binding ability, directs a modestly enhanced response.

Macrophages are an important physiologic target of IFN-γ.^4,27  To evaluate the biologic activity of the SUMO-ylation–defective Stat1 mutants, primary Stat1^–/– BMMs were infected with retroviruses encoding wild-type Stat1, Stat1^K703R, or Stat1^E705A. Modest differences in the level of expression were noted that correlated with an infection efficiency, which ranged between 25% and 35% (not shown). First, these macrophages were assessed for a Stat1-dependent ability to induce NO production upon stimulation with IFN-γ plus LPS (Figure 7B, top panel).^26,27,40  Modest differences between the capacity of wild-type and mutant Stat1 to direct NO production appeared correlated with differences in the level of Stat1 expression. Next, the Stat1-dependent ability of IFN-γ to induce MHC II expression in Stat1 “transduced” (ie, GFP⁺) macrophages was evaluated by FACS (Figure 7B, bottom panel).²⁶  Again, no significant differences were observed between wild-type Stat1 and the mutants, suggesting that Stat1^K703R and Stat1^E705A direct wild-type patterns of IFN-γ–dependent responses in primary macrophages.

PIAS proteins were initially identified for their capacity to regulate STATs under conditions of overexpression.¹⁴  However, recent studies identifying PIAS proteins as SUMO E3 ligase(s) raised the possibility that STATs may be regulated through SUMO modification.^{19,34,37,38,41,42}  When STAT sequences were scanned for the SUMO consensus modification site, only 2 potential sties were identified in Stat1, but not in other STATs (Table 1). Biochemical studies identified lysine 703, adjacent to the activation tyrosine (ie, Y701), as the SUMO conjugation site. However, the inability to recover tyrosine-phosphorylated SUMO-Stat1 conjugates in the HEK-293T cells raised concern over whether a single Stat1 molecule could be simultaneously modified at both Y701 and K703 (Figure 1). Consistent with this, a tyrosine-phosphorylated Stat1 peptide was a poor substrate for in vitro SUMO modification (Figure 2).

To evaluate the potential physiologic significance of SUMO Stat1 conjugation, SUMO-ylation–defective Stat1 mutants were generated. Fortuitously, the 2 most critical residues in the SUMO consensus modification site, K703 and E705, are exposed, suggesting they could be mutated without perturbing Stat1 activity.^6,32  Consistent with this, expression of the first mutant, Stat1^K703R, in Stat1^–/– MEFs revealed a wild-type pattern of rapid IFN-γ–dependent activation. However, the activation of Stat1^K703R was associated with significantly enhanced DNA binding and a prolonged pattern of nuclear retention. This observation raised the possibility that Stat1 SUMO modification might normally serve to promote Stat1 signal decay (eg, by driving DNA dissociation and thereby promoting dephosphorylation). Yet, biochemical studies failed to demonstrate that tyrosine-phosphorylated Stat1, the predicted SUMO-ylation target, could be SUMO modified. A second finding that was difficult to reconcile with this model was that the whole population of SUMO-ylation–defective Stat1^K703R exhibited dramatically enhanced DNA binding activity, even though only minute quantities of Stat1-SUMO conjugates were detected in wild-type cells, even under conditions of overexpression (Figure 1).^23,32,33  The most significant data to undermine the notion that SUMO conjugation regulates Stat1 activity came from the characterization of the second SUMO-ylation–defective mutant, Stat1^E705A. This mutant was essentially indistinguishable from wild-type Stat1. It exhibited a wild-type pattern of activation, signal decay, DNA binding, and nuclear retention. More importantly, in vivo studies demonstrated that Stat1^E705A and wild-type Stat1 were equivalent in their ability to direct the expression of IFN-γ target genes and IFN-γ–dependent antiviral responses. Similar results were obtained when Stat1^E705A was expressed in primary Stat1^–/– macrophages. In sum, these observations provide strong evidence that SUMO modification does not play an important role in the regulation of Stat1 activity, either in fibroblasts or macrophages.

Upon completing our initial characterization of the Stat1 SUMO-ylation site, 2 other groups reported K703 as a Stat1 SUMO modification site.^32,33  Both found that Stat1^K703R exhibited relatively modest changes in Stat1^K703R-dependent transcriptional activity in primate cells. Subsequently, 1 of these groups confirmed our observation of enhanced IFN-γ–dependent DNA binding activity and nuclear retention with Stat1^K703R.²³  Analogous to our own observations, this correlated with a modest increase in the Stat1^K703R-dependent transcription of a reporter, as well as prolonged transcription of 3 endogenous genes (eg, GBP1, IRF1, and TAP1). A second SUMO-ylation–defective mutant, Stat1^E705A, was also evaluated and found to drive a modestly enhanced IFN-γ–dependent expression of a GAS-driven reporter gene (versus our finding of a wild-type response). However, companion DNA binding and nuclear retention studies on Stat1^E705A were not provided. We speculate that the modest differences between these published studies and ours reside in the cell type used. Our study used Stat1-deficient primary cells or their derivatives, and not tumor cells that had undergone extensive mutagenesis, as is the case with U3A cells.⁴³  Moreover, our study used a pool of stably “transduced” cells, rather than a limited number clonally selected cell lines. (Of note, in our hands the pattern of gene expression varied considerably among U3A clones; data not shown.) Finally, our studies demonstrated that IFN-γ–dependent biologic responses, including antiviral activity, NO production, and target gene expression, were not significantly perturbed in SUMO-ylation–defective Stat1 mutants.

Although the unusual properties of Stat1^K703R could easily be exploited to argue that Stat1 activity is regulated by SUMO conjugation of K703, we believe that our data provide compelling evidence to the contrary. Notably, even under idealized reaction conditions, with purified Stat1, only a modest fraction of Stat1 was SUMO modified (Figure 2). This contrasted the fully penetrant-enhanced DNA binding activity of Stat1^K703R, suggesting that this mutation causes a structural perturbation to Stat1 dimer that stabilizes DNA binding. Consistent with this, K703 residue lies in a critical location of the Stat1-Stat1 dimerization interface. More significantly, a second SUMO-ylation–defective Stat1 mutant fails to exhibit this phenotype, providing compelling evidence that the enhanced DNA binding is unrelated to a potential loss in the ability to be SUMO modified. Surprisingly however, the enhanced DNA binding activity of Stat1^K703R correlated with relatively meek changes in biologic response, including a modestly enhanced antiviral activity at the lowest doses of IFN-γ and a lack of differences in NO production. The enhanced antiviral activity suggests that increased DNA binding activity may compensate for low levels of activation (ie, when active Stat1-Stat1 homodimers are rate limiting). Yet, at more standard doses of IFN-γ, this advantage is lost. The lack of correlation between the enhanced DNA binding activity of Stat1^K703R and target gene expression was surprising. This may suggest that Stat1 plays a more important role in initiating transcription (ie, an “on switch”) than in regulating the duration of a transcriptional response.⁴⁴  Moreover, the K703R mutation may impede recruitment of transcriptional cofactors, yielding an initially delayed response in target gene expression. Additional point mutations of K703 in Stat1, and closely related Stat3, will be important in exploring these possibilities.

We would like to thank Dr Jutta Braunstein for purified preparations of recombinant Stat1 and Stat3, and Dr Steve Goff for reagents and advice.