HSF2 is modified by SUMO-2/3. A. Cos7 cells were transfected with FLAG-HSF2 (wild type [WT]) or FLAG-HSF2 K82R (K82R). HSF2 was immunoprecipitated (IP) using an αFLAG antibody, and samples were analyzed by Western blotting (WB) with antibodies against HSF2, SUMO-2/3, or SUMO-1. Arrowheads denote the HSF2-sumoylated species recognized by αHSF2 and αSUMO-2/3 antibodies. Hsc70 was used as a loading control. B. The loop of HSF2 is sumoylated when inserted into HSF1. Cos7 cells were transfected with plasmids encoding HSF1 (WT) or HSF1Loop2 (1L2) together with GFP-SUMO-1 (+) or empty plasmid (−). Cells were heat shocked at 42°C for 15 min (HS) or maintained at 37°C (C). Cell extracts were subsequently analyzed by Western blotting with αHSF1 and αHsc70 antibodies. The sumoylated species are indicated on the left. The decrease in mobility of HSF1 in heat-shocked samples is due to hyperphosphorylation (39). Notably, coexpression of HSF1 and SUMO-1 in Cos7 cells led to constitutive HSF1 sumoylation, in contrast to the heat-inducible sumoylation previously observed in other cell types (13, 15). C. The loop of HSF1 is not sumoylated when inserted into HSF2. Cos7 cells were transfected with HSF2 (WT), HSF2Loop1 (2L1), or 2L1 mutants together with GFP-SUMO-1 (+) or empty plasmid (−). Cell extracts were analyzed by Western blotting with αHSF2 and αHsc70 antibodies. Notably, SUMO coexpression led to increased total levels of HSF2, which is likely due to the inhibitory effect of sumoylation on HSF2 DNA binding (see Fig. 4).

Ubc9 determines the differential sumoylation of the HSF1 and HSF2 loops. A. HSF2 and HSF2Loop1 interact equally well with Ubc9. In vitro-translated HSF2 proteins were incubated with recombinant GST or GST-Ubc9 as indicated. The ΨKXE-motif was disrupted by the I81S and E48A mutations. Bound fractions were released with denaturing buffer and subjected to Western blotting (WB) with an αHSF2 antibody. Equal amounts of GST and GST-Ubc9 were determined by Coomassie blue staining. WT, wild type. B. HSF2Loop1 is poorly sumoylated in vitro. Recombinant HSF2 and HSF2Loop1 were sumoylated in vitro using recombinant GST-Sae1/2 (E1), GST-Ubc9 (E2), and GST-SUMO-1. For controls, reaction buffer (−) was added instead of GST-Sae1/2 (+). Reactions were stopped by addition of denaturing buffer and analyzed by Western blotting using an αHSF2 antibody. C. The structural context determines the differential sumoylation of the HSF1 and HSF2 loops. GST-tagged peptides containing the loop of HSF1 or HSF2 were sumoylated in vitro as for panel B. Reactions were stopped by addition of denaturing buffer and analyzed by Western blotting using a SUMO-1-specific antibody. Equal input levels were determined by reblotting the membrane with an αGST antibody.

SUMO modification of the HSF2 loop is regulated by residues neighboring the consensus site. A. Schematic presentation of the nonconserved regions between the HSF1 and HSF2 loops. Mutations I, II, and III were introduced at the shaded sites, in which the amino acids of HSF2 were replaced with the corresponding amino acids of HSF1. B. Mutation of the GPV motif (mutation III) reduces HSF2 loop sumoylation. Extracts from cells transfected with wild-type HSF2 (WT) or HSF2 mutants containing HSF1-specific residues (I, II, and III), together with His-HA-SUMO-2 (+) or empty plasmid (−), were analyzed by Western blotting (WB) using an αHSF2 antibody. Equal protein input was determined by Hsc70 levels. C. Both the deletion of G87 and a P88D mutation reduce the sumoylation of the HSF2 loop. The experiment was done as for panel B. D. Close-up depiction of the interface between the ΨKXE motif from HSF1 (silver) or HSF2 (gold) and Ubc9 (blue). In comparison with HSF2, the presence of proline in HSF1 would lead to a much stronger hydrogen bond with Y87 of Ubc9 (the distance is smaller, 2.6 versus 3.3 Å, and the angle of bonding is ideal, nearly 180°). Such tight binding might prevent Y87 from altering its position as part of a catalytic mechanism and slow the release of sumoylated product. The effect of proline on its own would not be expected to completely prevent catalysis in the case of HSF1, since proline would not affect the relative positioning of the lysine and glutamate of the SUMO consensus motif. E. The region C terminal to the motif in HSF1 relative to HSF2 would limit the ability of the loop to place the sumoylation motif in an appropriate binding position. Here, the HSF model structures were superimposed over the elements of regular secondary structure of the DBDs but not over the loop regions, in order to reveal any relative displacements of the sumoylation motif. As a consequence of the unique structural properties of G87 and P88, in HSF2 two separate strong hydrogen bonds would be formed between main-chain atoms of V89 and V38, connecting the loop region to the final β-strand of the DBD sheet, stabilizing the structure, and facilitating better coordination of the VKQE tetrapeptide within the active site of Ubc9. In comparison, these strong hydrogen bonds would be replaced in HSF1 by a single, long and weak interaction between the main-chain nitrogen of E98 and oxygen atom of V46. Thus, in HSF1, the C-terminal end of the loop would be more detached from the β-sheet and more flexible than in HSF2, hampering Ubc9 from recognizing and docking VKPE of the loop to the active site of Ubc9. Due to the sequence differences C terminal to the sumoylation motif, in the modeled structures the nonsumoylated lysine of HSF1 is displaced by >4 Å with respect to the sumoylated lysine of HSF2.

Sumoylation of K82 inhibits the DNA binding of HSF2. A. EMSA performed on extracts from Cos7 cells transfected with the indicated plasmids (upper panel). The HSF2-HSE complex is indicated on the right. Input HSF2 levels were analyzed by Western blotting (WB) with an αHSF2 antibody (middle panel), and the αHsc70 antibody was used as a loading control (lower panel). WT, wild type. B. EMSA performed on reticulocyte lysate-translated WT or K82R HSF2 after sumoylation in vitro in the presence (+) or absence (−) of Ubc9 (upper panel). Aliquots of the sumoylation reactions were incubated at RT for 20 min and analyzed by Western blotting with an αHSF2 antibody (lower panel). C. Pull-down assay using an oligonucleotide containing three inverted NGAAN repeats (H) or a scrambled control (S). Extracts from Cos7 cells transfected with the indicated plasmids were incubated with an (NGAAN)₃ oligonucleotide. After the binding reaction, a flowthrough sample was taken and unbound proteins were removed by washing. The bound fraction was released by addition of denaturing buffer and analyzed by Western blotting using an αHSF2 antibody (left panel). Equal stoichiometry of sumoylated and nonsumoylated HSF2 species in the beginning (input) and end (flowthrough) of the binding reaction is shown in the right panel. Hsc70 is shown for equal loading.

Sumoylation on K82 does not interfere with HSF2 trimer formation. Cos7 cells were transfected with GFP-SUMO-1 together with empty plasmid or the indicated variants of Myc-HSF2 and FLAG-HSF2. FLAG-HSF2 was immunoprecipitated (IP) with antibodies against the FLAG epitope, and bound Myc-HSF2 was released by addition of denaturing buffer and detected by Western blotting (WB) using antibodies against the Myc epitope. Immunoprecipitated material is shown in the left panel and the input is shown in the right panel. Sumoylated and nonsumoylated Myc-HSF2 bind to FLAG-HSF2 with similar efficiency. Hsc70 was used as a loading control. WT, wild type.