To test whether stress conditions in general could increase the rate of whole chromosomal instability, we exposed haploid yeast cells to chemicals inducing various types of pleiotropic stress (Supplementary Table 1) for 12-14 hours and quantified chromosome loss rate by using the selection-neutral, chromosome fragment (CF)-based colony color assay (Fig. 1a, Supplementary Fig. 2; Supplementary Information) ¹¹. This initial screen revealed that many stress conditions, including hydrogen peroxide (oxidative stress), cycloheximide (translational stress), tunicamycin (ER stress), etc., elevated the chromosome loss rate to a level similar to that caused by benomyl, a microtubule inhibitor (Fig. 1a). Surprisingly, radicicol, an Hsp90 inhibitor ¹², was by far the most effective CIN inducer: the chromosome loss rate (7.4×10^-2/cell division) was hundreds of times above the control (2×10^-4/cell division), even at a radicicol concentration (10 μg/ml or 27 μM) with only minor effect on growth (Fig. 1a, Supplementary Fig. 3). Quantitative PCR (qPCR) confirmed that red colonies induced by radicicol had lost the whole CF (Supplementary Fig. 4a). Two of the 13 tested red colonies were confirmed to have also gained chromosome (Chr) X or Chr XI (Supplementary Fig. 4b, c).

A similar aneuploidy-inducing effect was also observed with macbecin II, a structurally distinct Hsp90 inhibitor (Fig.1b) ¹³. Deletion of one copy of Hsp90 genes, HSP82 showed enhanced CF loss compared to the wild type in the presence of radicicol or macbecin II (Fig.1b). Interestingly, deletion of STI1, the yeast homolog of mammalian Hop and a co-chaperone of Hsp90, resulted in significantly elevated CIN even at a concentration of radicicol too low to induce CIN on its own (Fig. 1b, Supplementary Fig. 5a). Heat is a common environmental stress known to tax Hsp90 function ¹⁴. Heat-shock for 90 seconds at 50.9°C induced subsequent CF loss at a rate comparable to that by pharmacological inhibition of Hsp90 (Fig. 1a). These results confirmed that Hsp90 stress is a potent inducer of aneuploidy. Hsp90 chaperon complexes are crucial facilitators of many cellular functions¹⁵. Previous biochemical studies suggested that Hsp90 is important for the activation of Ctf13 and assembly of the CBF3 inner kinetochore complex ³. Most CBF3 complex components, as well as the two co-chaperones involved in Ctf13 activation, showed haploinsufficiency toward radicicol (Supplementary Fig. 5b). Radicicol disrupted the kinetochore localization of Cep3 but had less effect on Ndc10, thus altering the stoichiometry of CBF3 complex at the kinetochore (Fig. 1c, Supplementary Fig. 5c, d, e). In addition to the CBF3 complex, Hsp90 interacts with several other pathways that could affect chromosome transmission fidelity, including the spindle assembly checkpoint ¹⁶ (see below).

Hsp90 taxation has previous been proposed to impact evolution by releasing phenotypic variation from pre-stored genetic diversity in the population and by transposon mobilization ^15,17. Does Hsp90 inhibition also promote adaptation through induction of aneuploidy? As a first test, a diploid strain was grown in the presence of high concentration of radicicol and 3 largest radicicol-resistant (Rad^r) colonies were selected and reconfirmed (Fig. 2a, Supplementary Figure 6 and Supplementary Information). Karyotyping revealed that all 3 Rad^r colonies were aneuploid with a dominant karyotype feature: all 3 Rad^r colonies, which adapted independently, contained one or two additional copies of Chr XV (Fig.2a). A haploid Chr XV disomy strain, generated by genetic manipulation ¹⁰, also showed strong resistance to radicicol (Fig. 2b). A previous genome-wide screen identified a set of genes exhibiting haploinsufficiency toward macbecin II, among which 2 of the top genes are located on Chr XV: STI1 and PDR5, a pleiotropic drug pump ¹⁶. We deleted a single copy of STI1 or PDR5 gene from Rad^r colony 3, trisomy for Chr XV. Growth measurements showed that either deletion abolished more than 50% of the growth rate gained by Chr XV trisomy over diploid in the presence of radicicol (Fig. 2c). A single copy of STI1 and/or PDR5 was then introduced into the parental diploid strain. An extra copy of each gene mildly but significantly increased radicicol resistance, but their combination drastically improved radicicol resistance (Fig. 2c, Supplementary Fig. 7). These results indicate that Chr XV gain directly confers radicicol resistance through increased copy number of STI1 and PDR5, and possibly also other genes carried on this chromosome (e.g., SGT1).

We next tested if the karyotype diversity produced by Hsp90 stress-induced CIN could fuel adaptation to various other stress conditions. A karyotypically mosaic yeast cell population (~1/3 of the population were aneuploid with different karyotypes, Supplementary Fig. 8c) was generated by growing a diploid strain under moderate Hsp90 stress (20μg/ml Radicicol) for 2 days. This population was then tested for enhanced adaptability toward other stress conditions, including the presence of growth inhibiting concentrations of fluconazole, tunicamycin, or benomyl, over a control homogeneous euploid population (see experimental scheme in Supplementary Fig. 8a). The radicicol-pretreated population did not show any growth advantage over the control diploid (vehicle pre-treated) on drug-free plates (Fig. 3a, Ctrl). However, on each of the different drug-containing plates, the radicicol pre-treated populations demonstrated drastically enhanced colony viability and frequency to form large drug-resistant colonies than the vehicle-pretreated population (Fig. 3a, b, c).

Twenty one colonies were picked from the vehicle control plates bearing the radicicol-pretreated population, and out of these 12 were aneuploid, whereas none (0/9) from the control plate bearing the vehicle-pretreated population were aneuploid (Fig.3d, e, Supplementary Fig. 8d, e). The vast majority (17/18) of the large colonies karyotyped from the drug plates bearing the radicicol-pretreated population were aneuploid (Fig. 3e). The drug-resistant colonies from the vehicle-pretreated population were also aneuploid (Fig. 3d). Importantly, the aneuploid colonies resistant to the same drug showed obvious karyotypic commonalities and tend to cluster together based on karyotype similarity (Fig.3e, Supplementary Fig. 9). For example, 4 of the 5 aneuploid colonies from fluconazole plates karyotyped gained an extra copy of Chr VIII, which carries Erg11, encoding an ergostrol biosynthetic enzyme known to confer fluconazole resistance in Candida albicans ¹⁸. Losing a copy of Chr XVI is a predominant karyotype change among the tunicamycin-resistant colonies (seen in 10/12 karyotyped, Fig. 3d). Of the 12 benomyl-resistant colonies, 10 demonstrated karyotype clustering with 6 of them losing one Chr XII, but it appears that more than one karyotypic pattern could confer benomyl resistance. This however is consistent with our previous observation of phenotypic convergence of distinct karyotypic patterns ⁶. All the above common karyotype features were significantly (Mantel-Haenszel tests) enriched in drug resistant colonies but not the starting radicicol-pretreated population prior to selection on drug plates (Fig. 3d, e, Supplementary Fig. 8e), suggesting an association of specific karyotypes with resistance to certain drugs.

To further assess the selective advantage of aneuploidy and karyotype dynamics under varying stress levels, two single Chr XVI monosomy colonies from a tunicamycin plate were streaked on drug-free plates. Colonies of two distinct sizes emerged, with the small ones being predominant (Fig.4a). Karyotyping showed that the small colonies represented Chr XVI monosomy, whereas the rare large colonies had gained back the missing Chr XVI and returned to diploid (Fig 4b, Supplementary Fig. 10a). Tunicamycin resistance was tightly linked to Chr XVI monosomy: all of the small colonies were tunicamycin resistant while the growth of the big colonies was abolished by tunicamycin (Fig 4c, Supplementary Fig. 10b, c, d). This result shows that an adapted aneuploid population also has the potential to return to euploid state when the stress condition is attenuated, suggesting that aneuploidy is not only a readily accessible mutation with large phenotypic impacts but is also reversible.

Taken together, the above results demonstrated that stress-induced chromosome instability, leading to aneuploidy, is a mechanism of stress-induced mutagenesis in eukaryotes with high adaptive value to diverse perturbations (Supplementary Figure 1). By far Hsp90 inhibition is the most potent inducer of aneuploidy among the stress conditions tested. This may be due to a broad but critical involvement of Hsp90 in pathways governing chromosome transmission fidelity and cell division ¹⁶. For example, the mitotic checkpoint gene MAD2 is a genetic interaction hub sensitive to Hsp90 perturbation ¹⁶. MAD2 deletion was also sufficient to lead to rapid emergence of fluconazole-resistant colonies bearing an extra copy of Chr VIII (Supplementary Figure 11). As Mad2 requires the CBF3 complex for its activity at the kinetochore ¹⁹, the exceptionally high-level CIN induced by Hsp90 inhibitors may be explained by a combined effect of interference with both kinetochore assembly and the checkpoint monitoring spindle defects. It is presently unknown whether the other stress conditions induce CIN through similar or different cellular targets.

The Hsp90 chaperon complex specializes in modulating the stability and function of many important regulatory and structural proteins ¹⁶. As a result, Hsp90 acts as a capacitor facilitating evolutionary adaptation by unleashing the effects of pre-existing mutations when Hsp90 activity is taxed under mild stress ^14,15. Strong Hsp90 inhibition also induces phenotypic variation through transposon activation in Drosophila ¹⁷. The results presented in this work reveal a new role for Hsp90 in adaptive evolution - as the guardian of chromosomal stability, the inhibition of which could trigger de novo karyotypic diversity leading to rapid adaptation through aneuploidy. We note that our observed induction of aneuploidy required more potent Hsp90 inhibition than that required to reveal phenotypic effects of pre-existing mutations ¹⁴. As the function of Hsp90 chaperon complex in kinetochore assembly is conserved in mammalian species ^4,20, the Hsp90 stress-induced aneuploidy may be a mechanism of cellular adaptation affecting a wide range of organisms.

1

2

3

4

5

6