Selective functional inhibition of a tumor-derived p53 mutant by cytosolic chaperones identified using split-YFP in budding yeast

Abstract Life requires the oligomerization of individual proteins into higher-order assemblies. In order to form functional oligomers, monomers must adopt appropriate 3D structures. Molecular chaperones transiently bind nascent or misfolded proteins to promote proper folding. Single missense mutations frequently cause disease by perturbing folding despite chaperone engagement. A misfolded mutant capable of oligomerizing with wild-type proteins can dominantly poison oligomer function. We previously found evidence that human-disease-linked mutations in Saccharomyces cerevisiae septin proteins slow folding and attract chaperones, resulting in a kinetic delay in oligomerization that prevents the mutant from interfering with wild-type function. Here, we build upon our septin studies to develop a new approach for identifying chaperone interactions in living cells, and use it to expand our understanding of chaperone involvement, kinetic folding delays, and oligomerization in the recessive behavior of tumor-derived mutants of the tumor suppressor p53. We find evidence of increased binding of several cytosolic chaperones to a recessive, misfolding-prone mutant, p53(V272M). Similar to our septin results, chaperone overexpression inhibits the function of p53(V272M) with minimal effect on the wild type. Unlike mutant septins, p53(V272M) is not kinetically delayed under conditions in which it is functional. Instead, it interacts with wild-type p53 but this interaction is temperature sensitive. At high temperatures or upon chaperone overexpression, p53(V272M) is excluded from the nucleus and cannot function or perturb wild-type function. Hsp90 inhibition liberates mutant p53 to enter the nucleus. These findings provide new insights into the effects of missense mutations.


Introduction
Higher-order protein structures are essential for life. Protein monomers must be synthesized, correctly folded, and efficiently assembled into homo-or hetero-oligomers while avoiding inappropriate interaction with misfolded partners. Quality control at each step is important to ensure these processes are accurate and timely. Molecular chaperones aid quality control by assisting protein folding through cycles of client binding and release. In a survey of nearly 3000 human disease-causing missense mutant proteins $30% showed increased association with chaperones relative to the wild-type (WT) allele, and this value is likely an underestimate (Sahni et al. 2015). Thus although chaperone binding to WT clients promotes client folding to the native state, in the case of misfolded mutants chaperone engagement does not guarantee that release of the client in a form that is competent for function. Although sequestration of grossly misfolded proteins by chaperones is an established mechanism for maintaining cellular fitness (Escusa-Toret et al. 2013), the idea that persistent chaperone engagement may actually interfere with normal protein function has received much less attention.
Septins are highly conserved cytoskeletal proteins whose assembly into hetero-oligomers is important for various cellular processes (Mostowy and Cossart 2012;Woods and Gladfelter 2021). Septins were originally discovered in Saccharomyces cerevisiae as temperature-sensitive (TS) mutants (Hartwell 1971). We previously demonstrated that at low temperatures permissive for cellular function many TS septin mutants are kinetically delayed in their folding, and that this delay has consequences for oligomerization when mutant septin monomers compete with WT monomers (Johnson et al. 2015), a quality control phenomenon we termed kinetic allele discrimination in oligomerization (KADO; Schaefer et al. 2016). Exclusion of the mutant septin from septin hetero-oligomers ensures that the mutant allele is recessive and does not interfere with the function of the WT allele (Johnson et al. 2015). We noticed that the excluded mutant persisted in the cytosol complexed with non-septin proteins, which we identified as cytosolic chaperones (Johnson et al. 2015).
Deletion of genes encoding some of the same cytosolic chaperones perturbed KADO and allowed TS-mutant septins to incorporate into hetero-oligomers despite the presence of the WT (Johnson et al. 2015). Thus chaperone binding to the mutant septin contributed to the kinetic competition between the two alleles.
In the case of a mutant that is nonfunctional at all temperatures, chaperone deletion converted the otherwise recessive mutant septin to a dominant-lethal allele, presumably by allowing the mutant to incorporate into septin hetero-oligomers and "poison" septin function (Johnson et al. 2015). Finally, chaperone overexpression inhibited septin function in cells in which only the mutant septin allele was expressed, consistent with a model in which chaperone engagement sequesters the mutant protein and blocks its ability to assemble into functional heterooligomers (Johnson et al. 2015). One yeast septin mutant subject to KADO carries the same amino acid substitution as a human mutation that blocks septin hetero-oligomerization and prevents septin function in sperm (Kuo et al. 2012;Weems et al. 2014;Johnson et al. 2015). We wanted to understand more about which chaperones bind WT and mutant septins and the extent to which the mechanisms we discovered with mutant yeast septins apply to other disease-linked mutations in oligomeric proteins.
p53 is a transcription factor and tumor suppressor that is mutated in nearly 50% of human tumors (see the IARC p53 Database, http://p53.iarc.fr). WT p53 exists on a precipice of thermostability, where missense substitutions easily push it "over the edge" in terms of its ability to fold, bind DNA and activate transcription (Bullock et al. 1997). Tumor-derived p53 mutants vary in conformation and function. In "contact" mutants, DNA binding is specifically abrogated by substitutions in the DNA-binding region without global effects on p53 conformation, whereas in "conformational" mutants the entire DNA-binding domain (DBD) is misfolded. p53 mutants can also be categorized as dominant or recessive in terms of their inhibitory effect on the function of coexpressed WT p53. Although fungi lack p53, yeast has long been an attractive system for characterizing p53 function using reporters whose transcription depends on p53 oligomerization, nuclear localization, and DNA binding (Sharma et al. 2016;Billant et al. 2017). Yeast-based analysis of tumor-derived missense mutations in the p53 core domain found that most were dominant-negative, loss-of-function alleles (Dearth et al. 2007). The remaining onethird failed to interfere with WT function and this recessive class was enriched for TS mutants (Dearth et al. 2007). We wondered if, like the recessive TS septin mutants, kinetic delays in folding/oligomerization or persistent chaperone engagement play a role in this class of p53 mutants. Coupled with the absence in yeast of many factors that exert extrinsic regulation on p53 oligomerization and turnover (e.g., MDM2; Di Ventura et al. 2008), the high conservation of all major classes of cytosolic chaperones allows us to focus on effects of mutations on p53 folding, chaperone engagement, and transcriptional activation function.
In this study, we set out to determine how chaperone engagement and kinetic delays in folding contribute to the recessive behavior of a TS p53 mutant. Having previously used functional approaches to find chaperone roles in recessive behavior for septin mutants, we first use mutant septins to test a bimolecular fluorescence complementation (BiFC) screening method to visualize chaperone-client interactions in living cells. We then combine BiFC chaperone screening with genetic and pharmacological manipulation of chaperones to gain a mechanistic understanding of the contributions of protein misfolding and cytosolic chaperones to the ability of a tumor-derived p53 mutant to dominantly interfere with WT function.

Materials and methods
Yeast strains, media, plasmids, and genetic manipulations Yeast strains are listed in Supplementary Table S1 and were manipulated using standard techniques (Amberg et al. 2005). Oligonucleotides are listed in Supplementary Table S2 and plasmids along with construction details are listed in Supplementary  Table S3. Yeast cells were cultivated in liquid or solid (2% agar) plates of rich or synthetic (Drop Out Mix Minus various ingredients; United States Biological, Salem, MA) medium containing total 2% sugar (glucose, raffinose, or galactose) or 4% sugar (2% each raffinose and galactose), as needed to maintain plasmid selection and induce expression from the GAL1/10 promoter. The TH5 strain was propagated on rich or synthetic medium with 100 mg/ml dTMP (thymidine monophosphate, TCI America no. TCT0845), or on minimal medium containing essential amino acids adenine, histidine, and methionine . Drop out medium containing 5-fluoro-orotic acid (FOA; United States Biological) was prepared with 50 mg/ml uracil and 0.1% FOA, except for use at 37 where modified to 0.07% FOA at pH 4.6. Radicicol (AdiopGen no. AG-CN2-0021) was dissolved in DMSO to 10 mg/ml. The bacterial strain DH5a (Agilent Technologies, Santa Clara, CA) was used to propagate plasmids.
Genomic or plasmid DNA from yeast was isolated by resuspending cells in 250 ml of P1 (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 100 mg/ml RNAse A), adding 0.5-mm glass beads to displace an additional $250 ml of volume, and vortexing for 3 min before addition of 250 ml P2 (200 mM NaOH, 1% SDS) and 350 ml N3 (4.2 M guanidine-HCl, 0.9 M potassium acetate, pH 4.8). Lysates were clarified by centrifugation and applied to Zyppy plasmid DNA Miniprep columns (Zymo Research) and DNA was washed and eluted according to the manufacturer's instructions. Alternatively, DNA was precipitated from clarified lysates using sodium acetate and ethanol (Green and Sambrook 2016). Yeast transformation was performed using Frozen-EZ Yeast Transformation Kit II (Zymo Research, Irvine, CA). Bacterial transformation was performed using CaCl 2 -prepared competent bacterial cells. PCR was performed with Phusion or Q5 highfidelity polymerase (New England Biolabs, Ipswich, MA) according to the manufacturer's instructions. Site-directed mutagenesis was performed by Keyclone Technologies (San Marcos, CA).

Microscopy
All micrographs were captured with an EVOSfl (Advanced Microscopy Group, Bothell, WA) all-in-one microscope equipped with an Olympus (Tokyo, Japan) 60Â oil immersion objective and yellow fluorescent protein (YFP), Texas Red, and GFP filters. Cells from mid to late log cultures were either taken directly from liquid cultures, applied to glass slides and covered with a coverslip, or concentrated by centrifugation, spotted to agarose pads, and covered with a coverslip. For septin microscopy, LED intensity was 100% and exposure times ranged from 500 to 1500 ms. For p53 microscopy, LED intensity for YFP and Texas Red ranged from 80% to 100%, and exposure time ranged from 1000 to 2500 ms, together chosen to produce no more than 10% saturated signal. GFP imaging used 20-30% LED intensity and 30-60 ms exposure. For comparison of samples in an experiment all samples were analyzed using the same intensity and exposure time.

Septin-chaperone BiFC
To measure bud neck BiFC fluorescence ( Figure 1C) we used an existing macro (Weems and McMurray 2017) with FIJI software (Schindelin et al. 2012). To measure intracellular BiFC signal not focused on the bud neck, we used FIJI to adjust the brightness/ contrast on transmitted light images of fields of cells and then used the Threshold function to create a binary image with cell outlines. The "Fill Holes" process was used to fill cell outlines, and then this binary image was used as a mask to measure the integrated signal intensity and areas for the fluorescence image    Table 1 for full dataset. Chaperones are color-coded by family and chaperones from the same family are drawn as the same shape. For the multisubunit chaperones PFD and CCT, if a single subunit interacted with any septin, then the entire complex is indicated as interacting with septins. (B) Micrographs showing discrete fluorescence signals in diploid cells coexpressing Cdc10-V C or the D182N mutant derivative with Ydj1-V N or Cct6-V N . The fluorescence images were inverted to facilitate viewing. Red arrowheads point to bud neck signal. Strains were diploids made by mating YO1057 with H00385 or H00382. (C) Cellular BiFC signal away from bud necks was quantified for diploid cells coexpressing Cdc10-V C or Cdc10(D182N)-V C and the indicated V N -tagged chaperones. The number of cells analyzed for each genotype ranged from 113 to 460. Strains were diploids made by mating YO1057 or H06530 with H00382, H00392, H00390, H00398, H00384, H00394, H00385, or H00396. (D) Representative images of diploid cells (strain M-1622 mated with H00398 or H00385) carrying cdc3(G365R) at one allele of CDC3 and WT CDC3 at the other, a V N -tagged allele of the indicated chaperone at one allele of the chaperone gene locus and an untagged allele at the other, and a low-copy plasmid (G00521 or G00522) encoding Cdc3-V C or Cdc3(G365R)-V C . Plot at right shows quantification of bud neck fluorescence. Twenty cells were analyzed for each genotype. of the same field of cells. To avoid non-cell particles, measurements were restricted to particles with areas between 500 and 2000 square pixels and circularity 0.5-1.0. Occasional cells (<1%) with extremely bright fluorescence were presumed to be dead and were excluded.

p53-chaperone BiFC
Haploid BY4742 cells carrying relevant plasmids described in figure legends were mated to V N -fusion strains encoding a V Ntagged chaperone . Diploids were inoculated in liquid synthetic medium with 1.9% raffinose and 0.1% galactose lacking leucine and uracil and imaged after growing to OD 600 $1 at 30 . For each chaperone the series of WT p53, p53(V272M), and p53(R273H) were imaged in parallel. Micrographs were quantified for Venus and mKate2 signal (n ¼ 100 cells for each genotype per chaperone) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of Venus and mKate2 signals from the same cell location (the cytosol, the nucleus, or both depending on the location of BiFC interactions). Relevant macros are available on GitHub (https://github.com/michael-mcmurray/ Denney-et-al- G3-2021/).

p53-p53 BiFC
Haploid Y0134 cells carrying relevant plasmids described in figure legends were inoculated in liquid synthetic medium with 2% glucose lacking leucine and tryptophan and imaged after growing to OD 600 $0.5 at 37 . Micrographs were quantified for nuclear Venus and mKate2 signal (n ¼ 100 cells for each genotype) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of Venus and mKate2 signals from the same cell location. Cells were selected for quantification based on the presence of Venus signal, as not all cells with mKate2 signal had generated and matured Venus signal, but all cells with Venus signal had corresponding mKate2 signal.

p53 localization assays
For kinetics of p53 nuclear accumulation, haploid yJM3164 cells carrying relevant plasmids were inoculated from overnight liquid cultures grown at 24 in synthetic medium with 2% raffinose lacking tryptophan into fresh medium with 1.9% raffinose and 0.1% galactose and imaged every 30 min for 6 h at 24 , and finally again at 10 h. Micrographs were quantified for GFP and mKate2 signal (n ¼ 100 cells for each genotype per timepoint) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of GFP and mKate2 signals from the nucleus. Note that GFP was used here only to verify the location of the nucleus and GFP quantities were not analyzed further.
For steady-state levels of p53 at different temperatures, haploid yJM3164 cells carrying relevant plasmids described in figure legends were inoculated in liquid synthetic medium lacking leucine and imaged after growing to OD 600 $0.5 at 30 or 37 . Micrographs were quantified for GFP and mKate2 signal (n ¼ 100 cells for each genotype per temperature) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of GFP and mKate2 signals from the nucleus. Note that GFP was used here only to verify the location of the nucleus and GFP quantities were not analyzed further.
For p53 localization with chaperone overexpression, diploid yJM3164 Â yJM1838 cells carrying relevant plasmids described in figure legends were inoculated in liquid synthetic medium with 2% galactose and 2% raffinose but lacking leucine and tryptophan and imaged after growing to OD600$1 at 30 . Micrographs were quantified for GFP and mKate2 signal (n ¼ 100 cells per genotype) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of GFP and mKate2 signals from the nucleus and cytosol. Note that GFP was used here only to verify the location of the nucleus (and by exclusion, the cytosol) and GFP quantities were not analyzed further.
For p53 localization with Hsp90 inhibition, haploid yJM3164 cells carrying relevant plasmids described in figure legends were inoculated in liquid synthetic medium with 2% glucose lacking leucine and containing 5 mg/ml radicicol or equal volume DMSO and imaged after growing to OD 600 $0.5 at 37 . Micrographs were quantified for nuclear GFP and mKate2 signal (n ¼ 100 cells per genotype) using FIJI software (Schindelin et al. 2012) and a macro that allowed selection and quantification of GFP and mKate2 signals from the nucleus and cytosol. Note that GFP was used here only to verify the location of the nucleus (and by exclusion, the cytosol) and GFP quantities were not analyzed further.

Coimmunoprecipitation and immunoblotting
Twenty-five milliliter yeast cultures in liquid synthetic medium lacking leucine and tryptophan were grown to OD 600 $1 at 30 and pelleted. Pellets were washed with Lysis Buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, 1 mM DTT) plus protease inhibitors (Complete EDTA-free; 11 873 580 001, Roche) and frozen in liquid nitrogen. Pellets were resuspended in 300 ml ice-cold Lysis Buffer plus protease inhibitors and lysed using 100 ml glass beads and eight cycles of 30-s on/off vortexing at 4 . Lysates were clarified by centrifugation into new tubes. Clarified lysates were diluted 1:2.5 in Binding Buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.5 mM EDTA, 1 mM DTT) for a total volume of 700 ml "Input." 25 ml Chromotek GFP-nAb Magnetic Agarose beads (Chromotek, Germany) were washed three times with Binding Buffer and combined with 700 ml diluted lysate. Beads were incubated while tumbling end-over-end for 1 h at 4 . "Unbound" fraction was removed and beads washed once with Binding Buffer and twice with Wash Buffer (10 mM Tris-HCl pH 7.4, 500 mM NaCl, 0.5 mM EDTA, 1 mM DTT). Beads were resuspended and gently pipetted in 50 ml 2Â SDS-PAGE sample buffer plus 5% b-mercaptoethanol and heated at 95 for 5 min to generate the "Bound" fraction. Following SDS-PAGE, proteins were electrophoretically transferred to PVDF and exposed to mouse anti-p53 (RRID: AB_10986581) and rabbit anti-Pgk1 primary antibodies. Infrared-labeled antimouse (680 nm, RRID: AB_10854088) and antirabbit (800 nm, RRID: AB_10697505) secondary antibodies were applied before scanning on a Licor Odyssey device. Input represents 7% of the amount of total protein used for the amount of immunoprecipitated material loaded on the gel.
For all other immunoblots, cultures were grown to mid-log phase and pellets were washed with 10% trichloroacetic acid (TCA) and frozen in liquid nitrogen, then placed at À70 C. Cell pellets were thawed in 150 ll of 1.85 M NaOH and 7.4% b-mercaptoethanol. 150 ll of 50% TCA was added before incubation on ice for 10 min. Precipitated proteins were pelleted at 4 C at maximum speed for 10 min, washed twice with cold acetone (1 ml each time), and resuspended in 80 ll of 0.1 M Tris base, 5% SDS. After addition of SDS-PAGE sample buffer, the samples were boiled for 5 min before resolution by SDS-PAGE. Membrane processing was the same as described above and also utilized, where indicated, mouse anti-6xHis antibodies (UBPBio no. Y1011). Ponceau solution (0.5% w/v, 1% acetic acid), where indicated, was applied for 30 min, then washed with water and the membrane allowed to dry before imaging.

Choice of sample sizes and statistical analysis
Sample sizes (number of cells) for fluorescence microscopy were chosen based on our ability to detect changes in subcellular localization in our previous studies (Johnson et al. 2015Weems and McMurray 2017). Analysis of effect sizes and generation of effect size plots were performed using DABEST (data analysis with bootstrap-coupled estimation; Ho et al. 2019) via the online server at http://www.estimationstats.com.

BiFC as an approach to trap otherwise transient chaperone-client interactions
In principle, the chaperone-septin interactions we found using biochemical approaches (Johnson et al. 2015) could have occurred following cell lysis. To visualize chaperone-septin interactions in intact, living cells, we used a BiFC approach, also called split-YFP. Reconstitution of Venus, an YFP derivative, occurs when two nonfluorescent Venus fragments, V N and V C , are brought in close proximity. Fusion of V N and V C to two proteins that directly interact in vivo is typically sufficient to drive Venus reconstitution. Although the interactions driving its folding and chromophore maturation are non-covalent, Venus reconstitution is effectively irreversible in vivo. Consequently, BiFC can be used to generate a stable fluorescent signal from otherwise transient interactions in which the untagged proteins would normally dissociate (Hu et al. 2002). Chromophore maturation requires considerable additional time after Venus folding, up to 50 min (Kerppola 2008); hence, signal intensity over time also depends on the half-life of the "crosslinked" fusion proteins. Since chaperone-client interactions are transient by definition, BiFC is ideally suited for detecting such interactions in vivo.
Venus fragments are misfolded proteins that could act as "chaperone bait" independent of the protein to which they are fused. Indeed, multiple V C -tagged chaperones (Hsp12, Hsc82, Hsp82, Hsp104, Ssa1, Ssb1, Ssb2, Sec63, Sis1, and Zuo1) are known to generate BiFC signals when coexpressed with "free" V N (Kim et al. 2019). We reasoned that the smaller size of V C ($120 residues, $13.5 kDa) compared with V N ($190 residues, $21 kDa) might make it less prone to nonspecific chaperone interactions. We therefore mated haploid yeast cells expressing V C -tagged septins with haploid cells expressing V N -tagged chaperones and after culturing the resulting diploid cells at room temperature ($22 ) we examined them by epifluorescence microscopy. We tested 48 V N -tagged chaperones with all five V Ctagged mitotic septins (Cdc3, Cdc10, Cdc11, Cdc12, and Shs1). As shown in Figure 1 and summarized in Figure 1A and Table 1, we observed discrete fluorescence signals for multiple cytosolic chaperones. In most cases these signals were found in either of two general subcellular locations. Although some diffuse BiFC signal was also faintly visible (see below), we focused first on the much more obvious discrete signals.
The first location of BiFC signal, observed with numerous chaperones, resembled the typical localization observed for native septins: rings at the cortex ( Figure 1B, Supplementary Figure S1A, and Table 1). Cortical rings were seen either exclusively at the bud neck, where septin rings are normally found, or faintly at bud necks and more intensely at locations reminiscent of budsite-selection landmark proteins that persist at previous sites of budding (Supplementary Figure S1A). In unstressed cells, most fluorescently tagged cytosolic chaperones are found distributed diffusely throughout the cytosol (Huh et al. 2003). We think it is most likely that for the chaperones generating cortical ring fluorescence, BiFC events arose from transient septin-chaperone Coexpression of WT p53 fails to rescue the folding defects of the p53(V272M) mutant at high temperature. (A) Illustration of split-DHFR p53 construct. TH5 cells are dfr1D and proliferate only when exogenous dTMP is present or p53 folds well enough to reconstitute active DHFR. (B) Colony growth at 30 and 37 by fivefold serially diluted TH5 cells carrying plasmids encoding the indicated p53 split DHFR alleles spotted on solid synthetic medium selective for the plasmids and lacking dTMP. Plasmids were G00598, "DHFR-WT"; G00604, "DHFR-V272M"; G00599, "DHFR-Y220C"; G00605, "DHFR-R273H." (C) As in (B), but with cells also carrying a plasmid encoding WT p53 (pLS76) or an empty vector ("EV," pRS415) on solid medium selective for the plasmids and either containing "þ" or lacking "À" dTMP at 37 . (D) As in (C) except cells were grown in liquid medium without dTMP at 37 in a flat-bottom 96-well plate. Optical density at 600 nm was recorded at 5-min intervals over 21 h for 6 replicate cultures per genotype and plotted as mean (thick line) and upper and lower bounds of the 95% C.I. (thin lines).
interactions in the cytosol. The tagged chaperone was then forced to follow the tagged septin on its subsequent cortical journeys. We interpret BiFC signal at bud necks as representing "crosslinked" septin-V C -chaperone-V N complexes in which the septin incorporated into septin hetero-oligomers and then into septin filaments, with the tagged chaperone becoming tethered to the septin ring. Due to the delay between initial interaction and fluorophore maturation, by microscopy we see only "old" septin-chaperone "crosslinked" complexes that may have already been through multiple rounds of filament polymerization and depolymerization. The appearance of bright cortical rings at sites of prior budding for some chaperones may indicate that those specific septin-chaperone BiFC associations perturbed the disassembly of "old" septin rings that normally follows each cell division, such that new rings are dimmer because they preferentially incorporate "new" or untagged septins. The second type of discrete BiFC signal was unique to interactions between WT septins and subunits of the cytosolic chaperonin called CCT (chaperonin containing tailless complex polypeptide 1). Coexpression of Cct7-V N with Cdc3-V C Cdc10-V C , or Cdc12-V C generated discrete puncta in a subpopulation of cells (Supplementary Figure S1A and Table 1). We engineered haploid strains carrying Cct7-V N and Cdc3-V C and confirmed that the punctate BiFC signal was not unique to diploid cells (Supplementary Figure S1A). That these haploids were viable and of normal morphology indicates that the septin-Cct7 BiFC interaction did not perturb essential CCT functions, and that only some of the Cdc3-V C molecules underwent BiFC events,  and expressing Hsp82-V N , Sis1-V N , or Cct7-V N and carrying plasmids encoding the indicated p53-mKate2-V C alleles. Total p53 (mKate2) fluorescence images were false-colored blue and BiFC (Venus) fluorescence images were false-colored orange. Prior to overlay, brightness and contrast were adjusted equivalently for each set of images. Plasmids were pMAM110, "WT"; pMAM112, "R273H"; pMAM111, "V272M." Strains were diploids made by mating BY4742 cells carrying the plasmids with H00398, H06680, or H00383. (B) BiFC (Venus) signal normalized to total p53 (mKate2) signal in cells cultured at 30 carrying plasmids encoding the indicated p53-mKate2-V C alleles together with the indicated V N -tagged chaperone. Blue lines denote median values. Vertical lines group the samples imaged with the same settings (LED intensity and exposure time). The subcellular location where the BiFC interaction was quantified is indicated below. Plasmids were pMAM110, "WT"; pMAM112, "R273H"; pMAM111, "V272M." Strains were diploids made by mating BY4742 cells carrying the plasmids with H00399, "Hsc82"; H00388, "Ssa2"; H00383, "Cct7"; H00384, "Hsp104"; H00398, "Hsp82"; H00395, "Ssb1"; H00394, "Ssb2"; H06680, "Sis1"; H00385, "Ydj1"; H00396, "Zuo1"; or H06684, "Hsp12."
otherwise no Cdc3-V C would have been available to perform septin functions at the bud neck. In stressed cells, other chaperones (e.g., Hsp104) localize to punctate structures together with misfolded proteins, such as the IPOD (insoluble protein deposit) and/ or JUNQ (juxtanuclear quality control; Kaganovich et al. 2008), but the Cct7-septin BiFC puncta did not colocalize with Act1(E364K)-mCherry, a misfolded form of actin that marks these compartments (Supplementary Figure S1B). CCT subunit C termini project into the center of the chamber (Llorca et al. 1999), hence Cct7-V Nseptin-V C interactions likely occur when septins are inside, as opposed to interacting with the chamber surface. Although each individual CCT subunit is relatively small, the cage-shaped holoenzyme is large ($1 MDa). We interpret the appearance of puncta as evidence that following BiFC association the tagged septin remains tethered to the inside of the chamber, and is unavailable for incorporation into septin-based structures. CCT assemblies with septins tethered via BiFC may be unable to fold other proteins, but why they would accumulate abnormally in a single intracellular location is a mystery. We cannot exclude the alternative possibility that Cct7-V N interacts with septins prior to assembling into a 16-mer; "free" Cct7-V N could interact with septins in a way that, when tethered to the septin via the BiFC connection, perturbs septin folding and drives aggregation in a way that does not involve the CCT holoenzyme. The apparent ability of Cdc3-V C to frequently "escape" Cct7-V N interaction could mean that some Cdc3 never interacts with CCT that some Cdc3-CCT interactions are too transient for BiFC, and/or that Cdc3-CCT interactions do not always involve the Cct7 subunit. Individual CCT subunits make client-specific contacts. For example, actin directly contacts only five CCT subunits (Balchin et al. 2018). When CCT is dissociated into individual subunits, Shs1 copurifies with Cct6 but not with Cct3 (Dekker et al. 2008), suggestive of some specificity in CCT-septin interactions. Indeed, Cct7-V N interacted with Cdc3-V C , Cdc10-V C , and Cdc12-V C , but did not exhibit a BiFC signal with Shs1-V C or its paralog Cdc11-V C (Table 1). We did not detect punctate septin BiFC signal with V Ntagged Cct2, Cct3, Cct4, or Cct6 in cells cultured at room temperature; Cct1/Tcp1, Cct5, and Cct8 are not in the V N tag collection and thus were not tested. Another multisubunit chaperone complex, prefoldin, also displayed subunit-specific BiFC interactions with septins (Table 1).
Temperature affects both protein folding and the expression of most chaperones. To ask if our BiFC system was able to capture such changes, we repeated our analysis in cells cultured at 37 . The spectrum of chaperone interactions at this temperature was mostly unchanged, with a few notable exceptions. At 37 Cct2, Cct3, and Cct6 showed detectable signal, as discrete puncta similar to what we saw for Cct7 at room temperature (  Figure 6 Chaperone inhibition liberates p53 from cytosolic compartmentalization. (A) Ratio of nuclear to cytosolic mKate2 signal in yJM3164 cells carrying plasmids encoding the indicated alleles of p53-mKate2-Venus. Cells were grown to mid-log phase at 37 with 2.5 mg/ml radicicol or equivalent volume of DMSO prior to imaging. Blue lines denote median values. Vertical lines separate genotypes. Plasmids were pMAM85, "WT"; pMAM86, "V272M"; pMAM87, "R273H"; pMAM104, "V272M R273H." (B) RBy33 cells carrying plasmids encoding the indicated alleles of p53 were grown to mid-log phase at 30 and lysed. Proteins were separated by SDS-PAGE. After electrophoretic transfer, the membrane was probed for p53 and the cytosolic protein Pgk1. Plasmids were pLS76, "WT"; pMAM69, "V272M"; pMAM79, "R273H"; pMAM77, "V272M R273H." (C) Representative micrographs of vacuolar mKate2 signal in yJM3164 cells carrying plasmids encoding the indicated alleles of p53-mKate2-Venus and treated with 5 mg/ml radicicol, prepared as in (A). Nucleus (Htb2-mNeon) fluorescence images were false-colored blue and p53 (mKate2) fluorescence images were false-colored orange. Prior to overlay, brightness, and contrast were adjusted equivalently for each set of images. Plasmids were pMAM85, "WT"; pMAM86, "V272M"; pMAM87, "R273H"; pMAM104, "V272M R273H." decreased at 37 . Septin interactions with prefoldin subunits almost completely disappeared at 37 (Table 1). Interactions between some septins (e.g., Cdc10) and the Hsp40 Ydj1 were no longer detectable, and interactions with several cytosolic Hsp70family chaperones were notably reduced (Table 1). Since many other chaperone interactions, including Hsp70s, were unchanged, an intrinsic reduction in the efficiency of Venus BiFC at 37 cannot explain the loss of specific chaperone signals. Instead, we attribute these changes to effects of temperature on septin folding-exposing different chaperone-interacting sequencesand/or chaperone expression levels.

Unique interactions between cytosolic chaperones and mutant septins that are kinetically delayed in their folding
Having established that WT septins associate with cytosolic chaperones, we wished to test a prediction from our KADO observations: that a slow-folding mutant septin would display distinct chaperone interactions relative to the WT allele. We previously showed that the D182N substitution in the GTP-binding pocket of septin Cdc10 renders the mutant protein unable to compete with WT Cdc10 for incorporation into septin complexes, and increases association with the chaperones Hsp104 and Ydj1, according to cofractionation and coprecipitation using cellular lysates (Johnson et al. 2015). To compare chaperone interactions using BiFC, we introduced the D182N mutation into the Cdc10-V C strain and mated with strains expressing V N -tagged chaperones.
In no case did we observe fluorescence in septin rings, which presumably reflects KADO-mediated exclusion of the mutant septin from septin complexes, and instead for some chaperones intracellular fluorescence was observed in other locations, including diffusely throughout the cells or in puncta ( Figure 1B and data not shown). To compare chaperone interactions between WT Cdc10 and Cdc10(D182N), we measured intracellular fluorescence using cell outlines, which mostly ignored cortical septin ring signals in WT cells (Supplementary Figure S1C). Whereas WT Cdc10-V C generated robust signal with Gim3-V N , almost no fluorescence was detected with Cdc10(D182N)-V C (Figure 1, C and D). Due to a lack of discrete fluorescence signal, we had not previously called Gim5, another prefoldin subunit, as a septin interactor (Table 1), but quantification revealed diffuse intracellular fluorescence with Cdc10-V C that decreased drastically for Cdc10(D182N)-V C (Figure 1, C and D). Conversely, little to no BiFC signal was found in cells coexpressing Cdc10-V C and Cct6-V N (Figure 1, B-D), in agreement with our qualitative calls based on discrete signals (Table 1), but Cdc10(D182N)-V C generated much higher signals, often visible as multiple puncta per cell (Figure 1, B-D). BiFC signal decreased drastically for the cytosolic Hsp90 isoform Hsp82, but not the more highly expressed isoform Hsc82 (Supplementary Figure S1D). BiFC signal also decreased for Ssb2 and Zuo1 (Figure 1, C and D), which comprise an Hsp70-Hsp40 pair within the ribosome-associated complex (RAC) of chaperones that engage nascent polypeptides and promote cotranslational folding (Gautschi et al. 2002;Preissler and Deuerling 2012;Zhang et al. 2017). For the Hsp40 chaperone Ydj1 the change in BiFC signal was more complex. Two populations of cells, differing in the intensity of BiFC fluorescence, were apparent in cultures of each genotype, but the difference was exaggerated by the D182N mutation in Cdc10, with about half of the cells exhibiting almost no signal and the other half having signal higher than the median signal for WT Cdc10 (Figure 1, B-D). Since we measured integrated signal density, we considered that differences in cell size could indirectly alter signals between genotypes, but when we corrected for cell size the same changes were still apparent (Supplementary Figure S1E and data not shown). Finally, Hsp104 BiFC signal was higher with Cdc10(D182N) (Figure 1, C and D), Our interpretation of the literature regarding the major subcellular localization for the chaperone in question, with minor localizations given in parenthesis. recapitulating our published biochemical results (Johnson et al. 2015). To gauge statistical significance, effect size measurements for most of these differences are provided in Supplementary Figure S1F.
Encouragingly, many of the BiFC differences we saw linked directly to our previous functional studies. The prefoldin subunit Gim3 is required for efficient KADO of Cdc10(D182N), as Cdc10(D182N)-GFP localizes to septin rings in a subset of gim3D haploid cells coexpressing WT Cdc10 (Johnson et al. 2015). Ydj1 is also required for Cdc10(D182N) KADO and Ydj1 overexpression inhibits the proliferation of cdc10(D182N) cells (Johnson et al. 2015). Hsp104 is dispensable for septin KADO but Hsp104 overexpression inhibits the function of mutant septins subject to KADO (Johnson et al. 2015). We note that the decreases in Cdc10 interactions with Ydj1 and prefoldin subunits upon introduction of a mutation predicted to affect folding state are qualitatively similar to those observed with WT Cdc10 in cells cultured at 37 . These results are consistent with a model in which changes in the folding pathway of Cdc10(D182N) compared with WT Cdc10 correspond to changes in the identity and/or duration of the interactions made with cytosolic chaperones.
To ask if these observations with Cdc10 could be extended to another slow-folding mutant septin, we tested Cdc3(G365R), which we have previously shown is slower to incorporate into septin complexes even at a temperature permissive for function and in the absence of a WT allele of CDC3 (Schaefer et al. 2016). We used a haploid strain carrying an untagged copy of the cdc3(G365R) allele at the CDC3 locus and introduced a V C -tagged allele of cdc3(G365R) on a low-copy plasmid, or WT CDC3 as a control. We then mated this strain with haploid strains carrying selected V N -tagged chaperones. This strategy allowed the tagged mutant septin to compete with the substoichiometric molecules of untagged WT Cdc3 and successfully localize to septin rings, as evidenced by bud neck fluorescence in cells coexpressing Hsp82-V N ( Figure 1D). That Cdc3(G365R)-V C showed clear BiFC signal with Hsp82-V N suggests that the decrease in Cdc10(D182N)-Hsp82 signal (Figure 1, C and D) may be specific to that mutant septin. Examination of septin ring fluorescence in cells expressing V N -tagged Ydj1 revealed a decrease in BiFC signal for Cdc3(G365R)-V C compared with Cdc3-V C ( Figure 1D), consistent with our results for Cdc10(D182N) and the decrease in Ydj1 interaction with WT Cdc3 observed at 37 . In summary, among the changes in Cdc10-chaperone interactions observed by BiFC upon introduction of a mutation predicted to slow Cdc10 folding are the two chaperones (Gim3 and Ydj1) that we have already shown to be functionally required for septin KADO, and the two chaperones (Hsp104 and Ydj1) for which we previously found genetic and biochemical evidence of increased physical associations with the mutant septin (Johnson et al. 2015). We conclude that BiFC is a useful tool to identify candidate chaperones whose expression can influence the function of misfolded mutant proteins.
Coexpression of WT p53 fails to rescue the folding defects of the p53(V272M) mutant at high temperature Having used mutant septins to establish BiFC as a useful tool, we wanted to identify differential chaperone interactions with mutant alleles of p53. However, our KADO model does not require chaperone interactions (Schaefer et al. 2016). Hence we first addressed other mechanisms that could prevent certain p53 mutants from acting dominantly to inhibit the function of the WT.
It was previously proposed (Dearth et al. 2007) that the reason some misfolded p53 mutants are recessive is because their proper folding is restored via interaction with properly folded WT p53 molecules in the context of "mixed" hetero-tetramers. The V272M mutant adopts a non-native conformation in a TS manner (North et al. 2002;Blanden et al. 2020) and acts recessively when coexpressed with WT p53 in yeast cells harboring a URA3 reporter gene containing p53 binding sites in the promoter region (Brachmann et al. 1996;Dearth et al. 2007; see Figure 3). To ask if WT p53 is able to "chaperone" the proper folding of p53(V272M), we used a split-dihydrofolate reductase (split-DHFR) assay (Pittman et al. 2012). Here, p53 is fused between two fragments of DHFR and only folded p53 supports DHFR activity, manifested as proliferation by dfr1D yeast cells on medium lacking exogenous dTMP ( Figure 2A). As shown in Figure 2B, dfr1D cells expressing a split-DHFR construct with p53(V272M) or p53(Y220C), another recessive and TS mutant (Dearth et al. 2007), barely proliferated at 37 on medium lacking dTMP. Equivalent constructs with WT or p53(R273H), a contact mutant that folds normally but cannot interact with DNA (Bullock et al. 1997), supported faster colony growth at all temperatures ( Figure 2B). Coexpressing untagged WT p53 had no obvious effect when proliferation was assessed using colony growth on solid medium ( Figure 2C). We also used a liquid culture growth assay and here observed only a very slight effect for Y220C ( Figure 2D). Based on these data, we think folding rescue via interaction with WT p53 is unlikely to explain why the mutants do not act dominantly to interfere with WT function.

High temperature inhibits p53(V272M) nuclear localization and coassembly with WT p53
We next asked if kinetic delays during de novo synthesis and/or folding of p53(V272M) relative to WT p53 might prevent V272M mutant molecules from coassembling with WT molecules into dysfunctional oligomers, akin to the KADO phenomenon we observed for septins. Oligomerization buries a nuclear export signal and ensures nuclear p53 localization (Stommel et al. 1999). Similar to how we monitored the kinetics of septin synthesis, folding and assembly by asking how long it takes a fluorescentlytagged septin to reach the septin ring at the bud neck (Schaefer et al. 2016), we monitored nuclear mKate2 signal after induction of mKate2-tagged p53 or p53(V272M). Since steadystate nuclear levels of WT p53 and p53(V272M) were equivalent at 24 ( Figure 3A, 10-h timepoint and data not shown), we performed our experiments at this temperature. Our quantitative analysis demonstrated similar kinetics of nuclear accumulation for both WT and p53(V272M) ( Figure 3A). Effect sizes with associated 95% C.I. for these and other quantifications of p53 fluorescence can be found in Supplementary Figure S2. These results suggest that at low temperature mutant p53 is not significantly delayed relative to WT p53 in achieving the oligomerization-competent conformation required to reach the nucleus.
By contrast, nuclear localization of p53(V272M)-mKate2 was reduced at 30 and greatly reduced at 37 ( Figure 3B). Temperature had a much smaller effect, if any, on nuclear levels of WT p53 and the "contact" mutant R273H ( Figure 3B). These data point to a failure in nuclear localization as a major cause of p53(V272M) dysfunction at high temperature. The inability of V272M to dominantly interfere with WT p53 when the two alleles are coexpressed may reflect a reduced ability of V272M to coassemble with WT molecules into "mixed" oligomers wherein the mutant perturbs WT function. If WT p53 interacts poorly with p53(V272M) at 37 that would additionally explain the results of our split-DHFR experiments.
We next used BiFC to test for p53-p53 interactions in living cells, using p53-mKate2-V C and p53-mKate2-V N constructs previously validated in mammalian cells (Gaglia and Lahav 2014). We quantified nuclear signal for total p53 (mKate2 fluorescence) and for p53-p53 BiFC (Venus fluorescence) at 30 and 37 . Whereas total nuclear p53 and BiFC signals indicative of p53-p53 interactions were indistinguishable for WT-WT and WT-R273H at 37 , for WT-V272M both total nuclear p53 and BiFC signals were reduced ( Figure 3C). Since WT p53 also contributed to nuclear signal, this means that at 37 coexpression of the V272M mutant partially inhibited the nuclear localization of WT p53. By normalizing nuclear BiFC signal to total nuclear p53 signal, we found that the relative proportion of nuclear p53 present as "mixed" oligomers was much higher for WT-V272M than WT-WT or WT-R273H ( Figure 3C). These data indicate that at 37 only a small number of V272M molecules are able to interact with WT p53 and remain the nucleus, supporting a model in which at high temperature V272M mostly fails to fold into an oligomerizationcompetent conformation and thus cannot inhibit WT p53 function.
In previous studies, including during cell-free translation in vitro (Nicholls et al. 2002), p53 dimerized cotranslationally (Gaglia and Lahav 2014) such that tetramers assemble from dimers composed of two monomers encoded by the same mRNA. Thus distinct alleles do not "mix" in the dimer form but rather in the tetramer form only. We tried to assess the extent to which cotranslational p53 dimerization might contribute to the interaction (or lack thereof) between WT p53 and p53(V272M), but in our yeast system the mutants postulated to block oligomerization before the dimerization (L344P) or tetramerization (L344A) step (Gaglia et al. 2013;Gaglia and Lahav 2014) did not reliably prevent interaction between differentially tagged p53 alleles, particularly in a BiFC context (Supplementary Figure S3). Thus we were unable to distinguish between cotranslational and posttranslational mechanisms of dimerization or determine the point at which oligomerization fails in the V272M mutant.
The V272M mutation converts the dominant R273H mutant to a recessive form If p53(V272M) is recessive because it misfolds and fails to interact with WT p53, and p53(R273H) is dominant because it interacts with WT p53 to form dysfunctional tetramers, then introducing the V272M mutation should sequester the R273H mutant in the cytosol and render it recessive. Indeed, quantification of nuclear fluorescence revealed that nuclear localization of p53(V272M R273H) was diminished at 30 and greatly diminished at 37 , patterns indistinguishable from that of p53(V272M) ( Figure 3B). Like p53(R273H), when expressed alone p53(V272M R273H) was transcriptionally inactive at all temperatures tested (Figure 3, D and E). At 30 , both p53(R273H) and p53(V272M R273H) dominantly inhibited WT p53 function ( Figure 3F). However, at 37 p53(V272M R273H) was unable to inhibit WT function ( Figure 3F), consistent with our prediction. These data point to temperatureinduced misfolding and cytosolic sequestration as the major mechanisms by which the V272M mutation perturbs oligomerization and compromises p53 function.

Cytosolic chaperone-p53 interactions in living cells
Given the evidence that misfolding of p53 mutants is responsible for dysfunction at high temperatures (reviewed in Bullock and Fersht 2001) and drives prolonged associations with at least three kinds of cytosolic chaperones (Blagosklonny et al. 1996;Sepehrnia et al. 1996;Singh et al. 2010;Wiech et al. 2012), we used our BiFC approach to systematically identify p53-chaperone interactions. We predicted that the conformationally unstable p53(V272M) would have altered chaperone interactions relative to WT p53 or the "contact" mutant p53(R273H). We used 28 of the same V N -tagged chaperones as with septins, paired with the p53-mKate2-V C fusion from the p53-p53 BiFC experiments described above. (Note that the V C tag here is slightly larger than the V C tag used with septins.) We identified interactions between p53 and numerous cytosolic chaperones ( Figure 4A, Supplementary Figure S4A, and summarized in Table 2). To account for differential proteasomal degradation of p53 mutants (Singh et al. 2010), we used the mKate2 tag to normalize the p53-chaperone BiFC signal (Venus) relative to total p53 signal (mKate2), which revealed increased interactions between p53(V272M) and many chaperones relative to WT p53 and p53(R273H) ( Figure 4B). Several chaperones-Hsp82, Ssb1, Sis1, and Ydj1-demonstrated increases in p53(V272M) interaction that were obvious even before normalization (raw values in Supplementary Figure S4B).
With the exception of Cct7 and Hsp12, and Hsp104, which generated decreased and increased BiFC signal with p53(R273H) compared with WT, respectively, most chaperones interacted in similar ways with p53(R273H) and WT p53, whereas p53(V272M) displayed multiple distinct interactions ( Figure 4B and Supplementary Figure S4B). p53(R273H) specifically abolishes contact with DNA without global conformational consequences ). On the other hand, p53(V272M) causes misfolding and conformational instability (North et al. 2002;Blanden et al. 2020). These observations are consistent with a model in which chaperones recognize specific, exposed sequences in p53 that are largely buried in the native fold. The V272M mutation results in prolonged exposure of these sequences and thus more opportunities for cytosolic chaperone engagement.
In some cases, we also saw differences in BiFC signal localization. At 30 , the temperature at which we cultured cells prior to microscopic examination, the mKate2-V C -tagged p53 alleles differed from each other in their localization even when no V Ntagged chaperone was expressed. WT p53 and the R273H mutant were seen in both the cytosol and the nucleus, whereas the V272M mutant was found almost exclusively in the cytosol (Supplementary Figure S4A, "No BiFC Control"). These observations agree with prior cell fractionation experiments in mammalian cells, in which WT p53 and p53(R273H) were found almost exclusively in the nuclear fraction, regardless of temperature (Zhang et al. 1994) whereas p53(V272M) showed a distinct shift toward the cytosolic fraction even at low temperatures (Dearth et al. 2007). Most p53-chaperone BiFC signals were in the cytosol, either diffusely, as puncta, or a mixture of both signal types (Table 2). Hsp70 chaperones Ssb1 and Ssb2 generated diffuse cytosolic signal whereas Ssa2, Cct7, and Hsp104 generated exclusively punctate signals (Table 2, Figure 4A, and Supplementary Figure S4A). Multiple p53-Cct7 puncta were seen in some cells ( Figure 4A), clearly distinguishing them from septin-Cct7 puncta. Chaperones producing a mixture of both diffuse and punctate cytosolic signal include the Hsp90 isoforms Hsp82 and Hsc82 (with brighter and more numerous puncta for Hsc82) and the Hsp40 family members Sis1, Ydj1, and Zuo1 (Table 2, Figure 4A, and Supplementary Figure S4A). Sis1 and Hsp12 were unique in producing nuclear signal in addition to cytosolic fluorescence ( Figure 4A and Supplementary3 Figure S4A). Sis1 generated nuclear signal for WT and p53(R273) but essentially none for p53(V272M) ( Figure 4A).
By monitoring the mKate2 signal, we found several cases in which coexpression of a V N -tagged chaperone altered the localization of mKate2-V C -tagged p53 ( Figure 4A and Supplementary Figure S4A). These observations are consistent with a model in which Venus reconstitution trapped the p53 molecules in the location where the chaperone is normally found. Nuclear mKate2 signal for p53(V272M) in cells coexpressing Hsp12-V N is a particularly striking example (Supplementary Figure S4A). In all cases, the Venus signal colocalized with the p53 mKate2 signal in the "new" location, further arguing that BiFC events were responsible ( Figure 4A and Supplementary Figure S4A). We interpret these data as evidence that, just as in septin-chaperone BiFC, making permanent an otherwise transient interaction incites a localization "tug-of-war." For most septin-chaperone interactions, the septin wins; for all p53-chaperone interactions, the chaperone wins. Only in the case of a subset ($5%) of p53(V272M) cells did p53 localize as puncta in the absence of a coexpressed VN-tagged chaperone (not shown). When expressed in yeast, WT p53 forms cytoplasmic puncta in response to various stresses (Park et al. 2021). Accordingly, we suspect that the BiFC puncta arise from p53 misfolding and abnormal oligomerization driven by prolonged association with the tagged chaperone. The results described above point to a number of chaperones as compelling candidates to influence the function of mutant alleles of p53.

Overexpression of specific chaperones inhibits p53(V272M) function
Differences in chaperone engagement might explain functional differences between p53(V272M) and WT p53 if a chaperone sequesters the mutant protein in a nonfunctional state. We asked if the function of p53 could be aggravated by chaperone overexpression at a permissive temperature, focusing on the chaperones identified by our BiFC experiments. Plasmids encoding chaperones under control of a galactose-inducible promoter were introduced into a diploid URA3 reporter strain also carrying a plasmid encoding untagged WT p53 or p53(V272M) under control of a constitutive promoter. Various cytosolic chaperones impeded the transcriptional function of p53(V272M) and, to a generally lesser degree, WT p53 ( Figure 5A). Exceptions included Hsf1DN, a constitutively active version of the major heat shock transcriptional regulator (Park et al. 2006), where WT p53 and p53(V272M) behaved similarly. By contrast, Hsf1DN, like expression of specific individual chaperones, inhibited the proliferation of cells with the cdc10(D182N) septin mutation (Johnson et al. 2015). Hsp90 isoforms Hsp82 and Hsc82 slightly but equivalently impeded mutant p53(V272M) relative to WT p53 although the intensity of signal and number of puncta were increased for endogenous levels of Hsc82 in BiFC experiments. Both Hsp90 isoforms were expressed at similar levels, as assessed by immunoblotting ( Figure 5B).
Since the transcriptional activation function we monitor here requires nuclear p53, one way chaperone engagement might inhibit p53(V272M) function is via cytosolic sequestration. We quantified nuclear vs cytosolic p53(V272M)-mKate2-Venus signal in micrographs and found that in most cases chaperone overexpression decreased the nuclear/cytosolic ratio of p53(V272M), with the exception of Ydj1 where the ratio was preserved ( Figure 5C). Notably, in this imaging-based approach, cytosolic p53(V272M) levels sometimes also decreased along with nuclear levels relative to empty vector (Supplementary Figure S5, A and B), as illustrated for Sis1 ( Figure 5D). However, when we tested Sis1 overexpression Measured at 30 . "-," no discrete signal. " ," fluorescence signal appeared as diffuse cytosolic; "," punctate cytosolic signal; " ," both diffuse and punctate cytosolic signal; "*," nuclear signal. c (P), physical; (F), functional. by immunoblotting, we saw no clear difference in total p53(V272M) levels relative to the empty vector control ( Figure 5D). Our imaging-based quantification may underestimate total cytosolic levels, since we excluded what is likely vacuolar p53 signal seen in some chaperone-overexpressing cells (e.g., Sis1 and Zuo1; Figure 5E). We interpret these data as evidence that overexpression of specific chaperones sequesters p53(V272M) outside the nucleus and impedes transcriptional activation.
Overexpression of some chaperones inhibited colony growth in general, even when we did not select for p53 function, and this effect was independent of the p53 allele expressed ( Figure 5A). Overexpression of these same chaperones in cells not coexpressing p53 has no such effect (Sopko et al. 2006;Johnson et al. 2015). High-level p53 expression is toxic to yeast and is thought to involve p53 binding to endogenous yeast genes and thereby perturbing essential gene expression (Inga and Resnick 2001;Leão et al. 2013;Park et al. 2021). We noticed a trend in which the chaperones with stronger effect on nonselective growth (i.e., those that induced the most p53 "toxicity") had more p53(V272M) in the nucleus ( Figure 5, A and C and Supplementary Figure S5, A and B). Conversely, the chaperones that specifically inhibited p53(V272M) transcriptional activation function and most strongly decreased the levels of p53(V272M) in the nucleus (Hsp31, Ssa4, Hsp26, and Cct7) had little to no effect on nonselective growth ( Figure 5, A and C and Supplementary Figure S5, A and B). Although the details of the underlying mechanism are beyond the scope of this work, these results support the idea that p53 toxicity in yeast requires nuclear localization. However, our findings further suggest that canonical transcriptional activation function is dispensable for toxicity, and that whatever function is responsible can be modulated by chaperone expression.

Chaperone inhibition liberates p53 from cytosolic compartmentalization
If association with a specific chaperone mediates cytosolic sequestration, then inhibiting the ability of that chaperone to bind p53 should allow more p53 to reach the nucleus. Indeed, when we inhibited Hsp90 function using radicicol (Sharma and Masison 2011), nuclear levels and nuclear/cytosolic ratios increased for WT p53 and all variants (Supplementary Figure S5, C and D and Figure 6A). As assessed by immunoblotting, total p53 levels were not detectably changed between DMSO-and radicicol-treated cells ( Figure 6B). Reminiscent of chaperoneoverexpressing cells, we also observed what is likely vacuolar p53 signal in cells treated with radicicol but not with DMSO ( Figure  6C); this extranuclear signal was not included in our quantification of cytosolic p53 (Supplementary Figure S5, C and D). These findings are consistent with other studies with mammalian cells in which pharmacological Hsp90 inhibition increased nuclear localization of a different p53 mutant (Whitesell et al. 1998).
Notably, in that study Hsp90 inhibition did not improve the function of the mutant p53 (Whitesell et al. 1998). To ask if the liberation of nonfunctional mutant p53 from chaperone sequestration might increase the ability of the V272M mutant to dominantly interfere with nuclear p53 function at 37 , we used FOA to select against activity of the URA3 reporter, but saw only a very slight apparent increase in colony growth in the presence of radicicol, and this increase occurred regardless of whether or not the cells coexpressed p53(V272M) (Supplementary Figure  S6A). Considering that even WT p53 is conformationally labile and often samples nonfunctional conformations (Hainaut et al. 1995;Cañadillas et al. 2006), this result might suggest that Hsp90 inhibition allows molecules of WT p53 in nonfunctional conformations to translocate to the nucleus, where they cannot activate transcription but may dominantly interfere with the function of any properly folded molecules. To test if a colonycolor-based reporter might be more sensitive, we coexpressed WT p53 with p53(R273H) in a strain in which p53 activity promotes expression of the ADE2 gene (Supplementary Figure S6B). However, diminished color development at 37 precluded analysis of dominant inhibition (Supplementary Figure S6D). Thus the functional consequences of releasing p53 from cytosolic chaperone sequestration remain unclear.

Discussion
Chaperone binding to a misfolded protein is widely assumed to either promote proper folding or protect cells from adverse consequences of irreversibly misfolded proteins. Accordingly, chaperone upregulation or overexpression should rescue mutant protein function. However, in some cases chaperones "inappropriately" sequester otherwise functional proteins, such as the DF508 mutant allele of CFTR, the most common mutation causing human cystic fibrosis (Amaral 2006;Balch et al. 2011;Lopes-Pacheco et al. 2015). Conversely, some mutant proteins can fold well enough to hetero-oligomerize with WT partner proteins but carry substitutions that "poison" the function of the heterooligomer. Here, chaperone sequestration can render recessive an otherwise dominant-negative mutant. We aimed to developed tools for identifying chaperones that differentially engage misfolding-prone mutant proteins, and use them to test if chaperone association affects the function and dominant-negative effect of a misfolding-prone mutant of p53.
Our BiFC approach identified differential septin-chaperone interactions that matched results from our prior functional studies (Johnson et al. 2015). Only 2 of 11 intended negative controls, Sec63 and Pam18, generated detectable septin BiFC signal (Table  1). We suspect that, in addition to its role in translocating polypeptides across the ER membrane (Ng and Walter 1994), the Hsp40 family member Sec63 transiently associates with many cytosolic proteins (including septins), only some of which are ultimately threaded into the ER lumen. Consistent with this interpretation, Sec63-V C interacts with "free" V N (Kim et al. 2019). Hence, in retrospect Sec63 was not a good negative control. The mitochondrially localized chaperone Pam18 interacted only with Cdc3 and only at 37 (Table 1). Although interactions between Cdc12 and Pam18 and between Cdc3 and Hsp78, another mitochondrial chaperone, were reported in high-throughput genetic (Davierwala et al. 2005) and two-hybrid ) screens, we doubt that these interactions are biologically meaningful.
Although many cytosolic chaperones are highly abundant proteins thought to interact with a wide range of proteins, only some interacted with septins in our assay. If chaperones merely engaged the V C portion of the septin-V C fusions, and not the septins themselves, we might expect a much broader spectrum of cytosolic chaperone BiFC interactions. For example, the Hsp70 Ssa1 is the most abundant chaperone in the cytosol of unstressed yeast cells (Brownridge et al. 2013), but it did not interact with any septin at any temperature. Conversely, its much less abundant, 98%identical paralog, Ssa2, interacted with all five septins (Table 1). Other studies using the same Ssa1-V N strain were able to detect BiFC interactions with other proteins (Sung et al. 2013), demonstrating that the Ssa1-V N fusion is functional. Ssa1 and Ssa2 also have distinct functions in prion propagation (Allen et al. 2005) and the vacuolar import and degradation pathway of gluconeogenic enzyme regulation (Brown et al. 2000; Sharma and Masison 2011). As another example, the Hsp70 Ssz1 forms a ribosomeassociated complex with Zuo1 and the other Hsp70s Ssb1 or Ssb2, but Ssz1 appears to have lost its ability to bind nascent substrate polypeptides (Huang et al. 2005). In agreement with this idea, we found that Zuo1-V N and Ssb1-V N , but not Ssz1-V N , interacted with V C -tagged septins (Table 1). Thus, we are confident that our BiFC results reveal specific associations, not nonspecific interactions or promiscuous encounters.
A previous study using a trimolecular fluorescence complementation (TriFC) assay to identify septin-interacting proteins in living yeast cells specifically tested interactions with Hsp82 as a negative control, and found none (Finnigan et al. 2016). We suspect that the septin-Hsp82 interactions we see with BiFC reflect the relatively large sizes of the V C and V N tags, the linkers between them, and the C termini of the septins and chaperones to which they were fused, compared with the very small (20-21 residues) tags and linkers (18-33 residues) used in the TriFC system (Finnigan et al. 2016). The TriFC study demonstrated that the small tags and short linkers limit signal generation to interactions occurring within a short distance, such that only septins adjacent to each other in the septin hetero-oligomeric complex generate TriFC signal (Finnigan et al. 2016). In our BiFC system, all septins interact with all septins (our unpublished observations). Accordingly, Hsp82 could probably interact with any part of a septin protein and generate BiFC signal, but would likely need to bind nearby the N or C terminus in order to generate TriFC signal. Since our KADO model predicts that chaperones recognize septins with misfolded GTPase domains (Johnson et al. 2015), which are formed primarily by residues in the middle of septin proteins, using a split-fluorophore approach to identify septin-chaperone interactions may require larger tags and longer linkers.
Accordingly, the drop in specific chaperone BiFC signals we observed for mutant septins relative to their WT counterparts does not necessarily mean those chaperone-septin interactions no longer occurred. Differences in conformation or exposure of chaperone binding sites for a mutant septin may increase the physical distance between the C-terminal V C tag on the septin and the C-terminal V N tag on the chaperone beyond the BiFC threshold. These differences may explain diminished BiFC signals for chaperones whose deletions (Ydj1 and Gim3) perturb septin KADO, and whose overexpression (Ydj1) impedes mutant septin function.
The changes in chaperone interactions upon septin mutation were similar to changes for WT septins at high temperature (Table 1). Certain chaperones may recognize non-native conformations that WT proteins only sample at high temperature but mutant proteins readily adopt at low temperature. Recent work on mammalian Hsp90, for example, suggests that client binding depends on the combined probabilities of partial client unfolding, capture by Hsp70 of a transiently exposed site, and further unfolding events that expose adjacent client regions for capture by Hsp90 in complex with cochaperones (Wang et al. 2020). The mutations we examined could promote any or all of these events. We chose V272M as a representative example of a p53 mutant with global structural instability and temperature-aggravated misfolding, and we have no reason to suspect that other p53 mutants with similar structural defects (e.g., manifested as temperature sensitivity) would behave differently.
Alternatively, the presence of a mutant protein might itself alter chaperone expression in ways that mimic the effects of high temperature, such that a different repertoire of chaperones is available for interaction, independent of protein conformation per se. Indeed, expression of mutant p53 in yeast cells was sufficient to alter chaperone expression without an increase in temperature (Singh et al. 2010). The BiFC tags themselves may also alter chaperone expression: we detected BiFC signals with WT septins for chaperones whose expression in nonstress conditions is normally very low (e.g., Ssa3; Werner-Washburne et al. 1987, Table 1). The V C tag itself did not drive chaperone binding; however, since p53-V C did not interact with Ssa3-V N (Table 2).
Whereas chaperone binding at the septin-septin G interface and competition with other septins for interaction there is consistent with predictions from our KADO model for mutant septins, most known chaperone binding sites in WT p53 are found in the structurally labile DBD (Fourie et al. 1997;Mü ller et al. 2004), as opposed to the oligomerization domain. If the same sites mediate chaperone binding to p53(V272M), then it seems less likely that chaperone binding directly blocks oligomerization by mutant p53 and, since oligomerization buries the p53 nuclear export signal, indirectly promotes cytosolic sequestration. The three p53 nuclear localization signals are all located near the C terminus and outside the DBD (Shaulsky et al. 1990), hence direct occlusion of these sequences by chaperone binding is also an unlikely mechanism of cytosolic sequestration.
Instead, chaperone binding to the p53 DBD could inhibit nuclear localization of p53 in other ways. In mammalian cells, p53 nuclear localization is controlled by many factors not present in yeast, including the protein MDM2, intermediate filaments, and dynein-mediated transport along microtubules (reviewed in O'Brate and Giannakakou 2003). Nonetheless, the same NLS and NES sequences that control p53 nuclear localization in mammalian cells also control p53 nuclear localization when expressed in yeast (Godon et al. 2005). Binding of mammalian Hsp70-Hsp40 to the WT p53 DBD alters its conformation to disfavor DNA binding (Boysen et al. 2019), but an inability to bind DNA cannot explain the cytosolic sequestration of p53(V272M) because the contact mutant R273H does not have the same problems with nuclear retention ( Figure 3B). We suspect that cytosolic chaperone-p53 complexes are retained in the cytosol because chaperone binding stabilizes a conformation of p53 in which the NLS is unavailable for proper interaction with the nuclear import machinery. Indeed, experiments in mammalian cells with a different TS p53 mutant, C135V, suggested that at high temperature binding of cytosolic Hsp70 chaperones to the mutant p53 renders the NLS inaccessible without directly occluding it (Akakura et al. 2001). p53 expressed in yeast can adopt several forms with distinct localization patterns-both punctate and diffuse, and in the cytosol, vacuole, or nucleus-some of which can be stably propagated in prion-like ways, suggesting conformational changes in p53 drive the changes in localization (Park et al. 2021). Since most of the chaperones we overexpressed are normally found diffusely in the cytosol, we think that the distinct patterns of p53 mislocalization outside the nucleus upon overexpression of distinct chaperones (punctate, diffuse, likely vacuolar) may have more to do with the conformation of p53 resulting from chaperone interaction than with chaperone binding per se. Along the same lines, the p53 we saw localized to the presumptive vacuole upon Hsp90 inhibition probably represents p53 liberated from Hsp90 but captured by other chaperones and ultimately excluded from the nucleus.
Our findings that overexpression of any one of many distinct chaperones specifically inhibited the function of p53(V272M) suggest that cytosolic sequestration was sufficient but not necessary for functional inhibition. For example, Ydj1 overexpression inhibited p53(V272M) function but actually increased nuclear p53 levels ( Figure 5A, Supplementary Figure S5, A and B). Hence some conformations of p53 are presumably able to enter and remain in the nucleus but are unable to function there properly. From a broader perspective, these functional data with p53(V272M) add to our previous results with mutant septin function (Johnson et al. 2015) and run counter to the idea that elevated chaperone expression generally improves the folding and function of mutant proteins. Although chaperone "assistance" with mutant p53 folding and function has been suggested in studies with other mutants expressed in yeast (including, interestingly, R273H; Singh et al. 2010), our data demonstrate that overexpression of the same chaperone (Ssa2) was inhibitory to V272M function ( Figure 5A). This discrepancy may point to differences in how chaperone binding alters the folding of a contact mutant (R273H) vs a conformational mutant (V272M). The extent and context of chaperone "overexpression" may also be important: the effect of Ssa2 on p53(R273H) was observed not by directly overexpressing Ssa2 but by adding ethanol, which indirectly induced Ssa2 (Singh et al. 2010). Similarly, some of the chaperones that had a functional effect on p53(V272M) were negative for BiFC interaction (i.e., Ssa4, Hsp26, and Hsp31). One potential explanation for this discrepancy is the high GAL1/10-driven chaperone expression we used in the functional assays, compared with endogenous chaperone levels in BiFC.
Another key point of comparison between our present results and those of prior studies is the effect of chaperone inhibition/deletion. In our earlier septin studies, deleting specific chaperoneencoding genes liberated mutant septins from apparent sequestration and allowed them to oligomerize with other, WT septins (Johnson et al. 2015). We focused on Ydj1 and Gim3 because published positive genetic interactions between ydj1D or gim3D and TS septin alleles hinted at improved function of the mutant septin when the chaperone was removed (Costanzo et al. 2010). In fact, however, the double-mutant cells were no more functional than either single mutant; they were merely less dysfunctional than one would have expected if each mutation independently compromised proliferation (Costanzo et al. 2010). Our finding that the Hsp90 inhibitor radicicol promotes nuclear localization of mutant p53 is conceptually similar, if we interpret it as p53 liberation from cytosolic chaperone sequestration but without fixing the folding problems that compromise function.
Indeed, our observation that radicicol also drives WT and R273H molecules of p53 into the nucleus ( Figure 6) may explain the unexplained finding (Godon et al. 2005) that various forms of cellular stress (radiation, cadmium, and mercury) promote localization of p53 to the yeast nucleus. If these stressors (all of which are known to induce the heat shock response) generate misfolded or damaged proteins, those proteins may exceed the sequestration capacity of the cytosolic chaperone machinery, and liberate misfolded p53 molecules to enter the nucleus. Such a mechanism would be reminiscent of the way proteostatic stress liberates Hsf1 from cytosolic Hsp70-Hsp90 sequestration and allows Hsf1 to trimerize, accumulate in the nucleus and activate chaperone gene expression (Morimoto 1998). Not all ways of inhibiting chaperone sequestration have the same effect; however, deleting the gene encoding the small heat shock protein Hsp26 rescues the function of p53(R273H) expressed in yeast (Singh et al. 2010). Presumably, p53(R273H) is able to achieve a functional conformation without Hsp26 assistance, whereas Hsp90 contribution to p53 folding is indispensable.
Finally, our findings do not support a KADO model for the recessive behavior of p53(V272M). KADO predicts delayed arrival of mutant V272M molecules in the nucleus and delayed mixing with WT, together with a recessive phenotype at all temperatures. Instead we observed a thermodynamic blockade on the ability of the vast majority of V272M molecules to reach a conformational state competent for nuclear localization. Sparing of WT inhibition by p53(V272M) and p53(V272M R273H) was only achieved at 37 , a temperature where the mutant was thermodynamically sequestered in the cytosol, rather than simply kinetically delayed in its nuclear arrival. For mutant septins, a poorly defined "window of opportunity" for hetero-oligomerization in the G1 phase of the cell cycle enforces mutant exclusion by KADO (Schaefer et al. 2016). We suspected that cotranslational dimerization might represent a p53 "window of opportunity," whereby a slow-folding mutant would be forced to dimerize posttranslationally and miss the chance to form mutant dimers that could mix with WT dimers. However, based on our findings it is unclear if in yeast WT p53 dimerization is cotranslational, and there is likely no kinetic "deadline" by which a p53 molecule must fold in order to properly oligomerize. Taken together, this study highlights key differences between septin oligomerization and p53 oligomerization and the effects of TS mutations and chaperone binding, which help us understand the general principles underlying dominant and recessive mutant phenotypes. oligonucleotides used in this study. Supplementary Table S3 contains plasmids along with construction details used in this study. Supplementary Figure S1 illustrates septin-chaperone BiFC interactions and quantification along with effect size plots and associated 95% C.I. Supplementary Figure S2 contains effect size plots with associated 95% C.I. corresponding to various p53 measurements found in Figures 3, A-C Figure S4 illustrates positive p53-chaperone BiFC interactions not included in Figure 4A along with raw BiFC and total p53 values corresponding to data in Figure 4B. Supplementary Figure S5 contains raw nuclear and cytosolic p53 values, along with effect size plots and associated 95% C.I., corresponding to data in Figures 5C and 6A. Supplementary Figure S6 depicts URA3-and ADE2-based p53 reporters and their utility in detecting dominant effects of p53 mutants.

Conflicts of interest
None declared.