Elucidating the regulatory mechanism of Swi1 prion in global transcription and stress responses

Transcriptional regulators are prevalent among identified prions in Saccharomyces cerevisiae, however, it is unclear how prions affect genome-wide transcription. We show here that the prion ([SWI+]) and mutant (swi1∆) forms of Swi1, a subunit of the SWI/SNF chromatin-remodeling complex, confer dramatically distinct transcriptomic profiles. In [SWI+] cells, genes encoding for 34 transcription factors (TFs) and 24 Swi1-interacting proteins can undergo transcriptional modifications. Several TFs show enhanced aggregation in [SWI+] cells. Further analyses suggest that such alterations are key factors in specifying the transcriptomic signatures of [SWI+] cells. Interestingly, swi1∆ and [SWI+] impose distinct and oftentimes opposite effects on cellular functions. Translation-associated activities, in particular, are significantly reduced in swi1∆ cells. Although both swi1∆ and [SWI+] cells are similarly sensitive to thermal, osmotic and drought stresses, harmful, neutral or beneficial effects were observed for a panel of tested chemical stressors. Further analyses suggest that the environmental stress response (ESR) is mechanistically different between swi1∆ and [SWI+] cells—stress-inducible ESR (iESR) are repressed by [SWI+] but unchanged by swi1∆ while stress-repressible ESR (rESR) are induced by [SWI+] but repressed by swi1∆. Our work thus demonstrates primarily gain-of-function outcomes through transcriptomic modifications by [SWI+] and highlights a prion-mediated regulation of transcription and phenotypes in yeast.


Results
Distinct transcriptomic profiles of [SWI + ] and swi1∆ cells. To examine how [SWI + ] and swi1∆ differentially affect genome-wide gene expression, we performed RNA-seq experiments to compare the transcriptomes of a set of isogenic S288C strains-wt, [SWI + ], and swi1∆. All three strains contain a repaired, and thus, functional FLO8 gene that is required for expression of flocculin genes and exhibition of multicellular features 31 . Prior to RNA-seq experiments, we confirmed the prion statuses of Swi1 and Rnq1. Rnq1 is the protein determinant of [PIN + ] 42 , also known as [RNQ + ] 43 . The prion status of Rnq1 was examined because de novo formation of [RNQ + ] can be induced by [SWI + ] 35 . As shown in (Fig. S1a), the diffused fluorescent patterns of Rnq1GFP in all three strains indicate that none of them contained the [RNQ + ] prion, excluding the interference of [RNQ + ] to our RNA-seq results. The Swi1 prion state was verified by the aggregation status of Swi1-NQ-YFP, reduced ability to use non-glucose carbon sources, and lack of adhesive growth 16,31 (Fig. S1a,b). Total RNAs were isolated from triplicated samples and cDNA libraries were constructed after oligo dT-enrichment. Upon sequencing and aligning the reads to a reference yeast genome (S288C strain), we found that one sample of the [SWI + ] strain and one sample of the swi1∆ strain clustered a little more apart from the other two replicates though all triplicates of the three strains still visibly grouped together (Fig. S1c). Likely, the prion may have been partially lost for the departed [SWI + ] sample due to instability of [SWI + ] 32 and genetic alterations may have occurred for the departed swi1∆ sample as reported previously 44 during cultivation. To ensure the robustness of the sequencing quality, data from the two departed samples were excluded from our analyses. As shown in Fig. S1d, ~ 15-18 million reads, comparable to that of approximate 62-75 × yeast genome size, were collected for each sample and 97-99% of the reads were uniquely mapped to the genome, indicating a good quality of the sequencing. The included yeast isolates do not show ploidy changes (Fig. S1e). The sequencing quality is also justified by confirming that several previously reported genes including FLO1 and FLO11 that are downregulated in swi1∆ and [SWI + ] cells 31 were indeed similarly downregulated as shown in Fig. S2. These data support the trustworthiness of our RNAseq results.
The cutoff for calling differently expressed genes (DEGs) in this study was a fold-change > 2 and an adjusted p-value < 0.001. Using such a criterion, 782 and 1128 DEGs were collected for [SWI + ] and swi1∆ strains compared to wt, respectively (Fig. 1a, also see Table S1 for raw sequencing data). These DEGs distribute evenly to every chromosome without noticeable biases (Fig. 1b). Surprisingly, [SWI + ] and swi1∆ strains only share 263 DEGs-23.3% of swi1∆ DEGs or 33.6% of [SWI + ] DEGs (Table S1 and data not shown). Consistent with this notion, we observed 1562 DEGs (954 are activated and 608 are inhibited by [SWI + ]) when a pairwise comparison of [SWI + ] and swi1∆ strains was conducted (Fig. 1a,c).
We then performed a detailed analysis of the identified DEGs of [SWI + ] and swi1∆ obtained by comparing to the wt transcriptome (Fig. 1d). A total of 287 upregulated and 495 downregulated genes were found for [SWI + ]  www.nature.com/scientificreports/ cells, whereas 397 upregulated and 731 downregulated genes were found for swi1∆ cells, demonstrating that the deletion imposes a greater impact on the yeast transcriptome than [SWI + ] ( Fig. 1d and data not shown). Combining DEGs of [SWI + ] and swi1∆, we observed a total of 1647 individual DEGs. These genes can be grouped into 8 subgroups based on their regulatory patterns by [SWI + ] and/or swi1∆ (Fig. 1d), individual DEGs in each subgroup are dispersed evenly across the genome without obvious hotspots or biases for specific DNA regions (Fig. S3). Further investigation revealed that most of the DEG-transcripts are protein coding regions-open reading frames (ORFs) (Fig. 1e). In this regard, among the total 6 (Fig. 2a). The large number of TFs and their targets found in this analysis was unexpected. One possible explanation is that the recruitment of the SWI/SNF complex to its target promoters requires coordinated actions of certain TFs, and thus Swi1 regulated genes are often TF targets 47,48 . Nevertheless, the above findings encouraged us to search for individual TF genes that are transcriptionally regulated by [SWI + ] or swi1∆, and as a result, we found 34 and 31 such TF genes among DEGs from [SWI + ] or swi1∆ cells, respectively (Table S2). Further analysis suggests that TF genes are more significantly overrepresented within the prion DEGs, particularly DEGs downregulated by [SWI + ] (Fig. 2b). Interestingly, for TFs having overrepresented targets within [SWI + ] DEGs, 22 of them actually are transcriptionally modified by the prion, including 7 upregulated (MIG2, YPR015C, RDS1, BAS1, NRG2, RPI1, GCR1) and 15 downregulated (SUT1, XBP1,  HMS1, MGA1, CRZ1, GIS1, MIG1, RCS1, RLM1, RPN4, CAT8, ADR1, EDS1, GAL3, ZNF1) TF genes (Fig. 2a and  Table S2). Notably, MIG2, YPR015C, RDS1 are upregulated and SUT1, XBP1, HMS1, MGA1 are downregulated similarly in swi1∆ cells. 765 targets of the 22 TFs were identified from the [SWI + ] DEGs-281 activated and 484 inhibited (Fig. 2c, also see regulatory details in Fig. S4 and Table S2). As activators, the upregulation of 7 TF genes can explain the enhanced transcription of 144 DEGs; and as inhibitors, their upregulation can explain the downregulation of 395 DEGs in [SWI + ] cells. Similarly, the upregulation of 250 genes and downregulation of 462 genes in [SWI + ] cells are interpretable by activation and inhibition activities of the 15 downregulated TFs, respectively (Fig. 2c). After removal of the redundant targets, the activation of 262 DEGs (93.2% out of the 281 total upregulated DEG targets) and the inhibition of 482 DEGs (99.6% of the total 484 downregulated DEG targets) are interpretable by transcriptional alterations of the 22 TF genes (Fig. 2c). The interpretable percentage is not 100%, perhaps due to that some DEGs can be regulated by multiple TFs that can act as activators or inhibitors depending on their targets. For example, Bas1 was predicted to be one of the most versatile upregulated TFs, which can explain the downregulation of 327 DEGs to which it serves as an inhibitor and for the upregulation of 96 DEGs to which it functions as an activator (Table 1 and Table S2). In contrast, Rnp4 is one of the key TFs that is downregulated in [SWI + ] cells and can be used to explain the downregulation of 395 DEGs and the upregulation of 222 DEGs (Table 1 and Table S2) through activation and inhibition, respectively. The combined actions of Bas1 and Rnp4 may also alter the regulatory output of some targets co-regulated by the two TFs. Interestingly, EDS1 and ZNF1 are downregulated and were predicted having overrepresented targets uniquely in [SWI + ] but not in swi1∆ cells. As shown in Fig. 2d

Aggregation of TFs and DEGs in [SWI + ] cells.
Next, we tested if conformational changes of TFs also contribute to the observed DEGs in [SWI + ] cells. This test was inspired by our previous observation that besides Swi1, several additional transcriptional activators containing an asparagine-rich region(s), including Mss11, Sap30 and Msn1, are significantly aggregated in [SWI + ] but not so in non-prion cells 31 . With focusing on TFs whose targets are significantly overrepresented within the [SWI + ] DEGs, we examined their aggregation/prionprone propensity. Utilizing two published algorithms, CamSol 49 and PrionW 50 , which were designed for identifying aggregation-prone and prion-like proteins, 30 TFs with an intrinsic variant score < 1.5 using CamSol or a pWALTZ cut-off of > 73.5 and N + Q > 20% using PrionW were selected for further study (Table S2). Initially, isogenic [SWI + ] and wt strains with individual TF genes endogenously tagged with GFP were used for the test. Unfortunately, we failed to see visible GFP foci for multiple TFs including Mss11, Sap30 and Msn1 that form amyloid-aggregates in [SWI + ] cells 31 . We speculated that the endogenous level of these TFs is too low to visualize their aggregation. Thus, we turned to express their GFP fusions from single-copy plasmids driven by a galactose-inducible GAL1 promoter. We observed that the aggregation of these fusion proteins indeed depended on their expression levels. With a galactose concentration of 2% (an overexpression condition), many of them formed foci in both [SWI + ] and wt non-prion cells with varying frequencies (data not shown). After testing different galactose concentrations and induction times, 0.02% galactose and up to 8 h of induction were used in the formal experiments to avoid unnecessary overproduction. Under such conditions, Swi1-NQ-YFP (positive control) aggregated in [SWI + ] but not in non-prion wt cells and YFP (negative control) did not aggregate in either type of cells (Fig. 3a,b). We observed dramatic increases in aggregation of four TFs-Mot3, Rds1, Mac1 and Rpi1 in [SWI + ] cells while their aggregation frequency in wt (non-prion) cells was kept at minimal levels except for Mac1, which also significantly aggregated in non-prion cells but with a considerably lesser degree when compared to that in [SWI + ] cells (Fig. 3a,c). The observed foci (aggregates) may reflect a change in protein localization, accumulation of amorphous aggregates, or formation of amyloids. Regardless of their nature, one consequence of such aggregations is the reduction of the functionality of these TFs and thus alterations of their target genes.
The four TFs are linked to 239 targets (168 downregulated/71 upregulated) among prion DEGs (Table S2). Nevertheless, only 86 (51.2% out of the 168 targets) downregulated genes and 34 (47.9% out the 71 targets) upregulated genes in [SWI + ] cells are interpretable by the possible reduced activities of the four TFs due to aggregation ( Fig. 3d and Table S2). Interestingly, Mot3, one of the four TFs, is the protein determinant of the  (Table S2). The aggregation of Rpi1 explains the reduced transcription of 5 and the increased transcription of 3 target genes. The aggregation of Mac1 is perhaps responsible for the reduced transcription of 13 genes and the increased expression of 3 genes.  Table S2 for details). To avoid unnecessary overproduction, 0.02% galactose was used to induce the expression of the indicated proteins from a single-copy plasmid. Representative images of the indicated YFP fusions are shown in panel (a), and quantitative results are shown in panel (b) (Swi1-NQ-YFP as positive control and YFP as negative control) and in (c) (for tested TFs). (d) Reduced activity of the 4 TFs due to aggregation explains the inhibition (upper) and activation (bottom) of [SWI + ] DEG targets when they function as activators and inhibitors, respectively. All TFs in the regulatory network are highlighted by blue. www.nature.com/scientificreports/ Similarly, the aggregation of Rds1 might be the cause of the reduced transcription of 6 and, the increased expression of 3 genes ( Fig. 3d and Table S2). Taken together, our data suggest that the aggregation of a group of TFs may be another contributory factor to our observed [SWI + ] transcriptomic profile.  Fig. 3d can be interpreted by a combined effect of the aggregation of Swi1 and the four TFs. We reasoned that the regulatory effect of Swi1 aggregation on these DEGs in [SWI + ] cells (with reduced Swi1 function) can be roughly estimated by their transcription in swi1∆ cells (in the absence of Swi1). Meanwhile, the regulatory effect of TF aggregation (reduction in function) on these DEGs can be predicted using YEASTRACT. As shown in Fig. 4a, the 120 DEGs can be divided into 6 subgroups based on their expression patterns in swi1∆ and [SWI + ] cells. For genes that are similarly regulated in [SWI + ] and swi1∆ cells, such a similarity can be explained by the fact that the aggregation of the four TFs and the SWI1 deletion have similar regulatory effects. For the other four subgroups of DEGs that are regulated differently by swi1∆ and [SWI + ], their transcriptional patterns can also be explained by a combined regulatory effect of the aggregation of Swi1 and the four TFs in [SWI + ] cells (Fig. 4a).

Contribution of Swi1 interaction proteins to DEGs in [SWI
In another analysis, we identified 325 swi1∆-regulated genes that are targets of the four TFs. Their transcription should be similarly modified in [SWI + ] cells if the reduced Swi1 activity were the only regulatory factor. We found that 93 such DEGs show similar regulatory patterns (78 inhibited and 15 activated) in swi1∆ and [SWI + ] cells. For these genes, the effect of insufficient Swi1 function might be the primary cause. For the other four subgroups of genes (total 232), their transcription profiles are different in swi1∆ and [SWI + ] cells. As shown in Fig. 4b, the aggregation effects of Swi1 and the TFs are variable, and whether or not a regulatory pattern in [SWI + ] cells is interpretable by the aggregation of Swi1 and the four TFs depends on individual genes. Overall, the transcriptional patterns may be explained by a combined effect of the compromised Swi1 function and/or the aggregation of TFs for 66.5% but not for the remaining 33.5% of DEG targets of the four TFs in [SWI + ] cells (Fig. 4b).
Given that the expression of a substantial number of [SWI + ] DEGs is not interpretable by the combined effects of Swi1 activity reduction and the TF aggregation, other alternative mechanisms may be involved. Interestingly, SWI1 gene expression is upregulated ~ twofold in [SWI + ] cells (Fig. 4c). There are 76 SWI1 regulators (e,f) The upregulation of RPI1 and RDS1 in [SWI + ] cells, which encode two aggregation-prone TFs (Rpi1 and Rds1), can be explained by transcriptional alterations of their regulators, especially those underlined TFs. E, RPI1 regulators, and F, RDS1 regulators. In panels of (a,b), ↑, activation; ↓, inhibition; −, no transcriptional change. In panels of (d-f), dotted green lines denote activation, dotted red lines denote inhibition, and dotted brown lines denote activation / or inhibition based on expression evidence from literature, and all TF genes are highlighted in blue in the shown regulatory networks. www.nature.com/scientificreports/ in the genome based on data retrieved from SGD (Fig. 4d). Among them, the expression of Bas1, Mig2, and Rpi1, which are Swi1 activators, is upregulated, while Rpn4, a Swi1 repressor, is downregulated in [SWI + ] cells (Fig. 4d left). These 4 altered TFs may form a regulatory network to result in SWI1 upregulation in [SWI + ] cells, and such an upregulation may partially alleviate and neutralize the effect of Swi1 aggregation and further alter the transcriptome of [SWI + ] cells. In addition, we also observed that the transcription of genes encoding Rds1 and Rpi1 that are prone to aggregate in [SWI + ] cells is increased (Table S2). This may be explained by an altered expression of several TFs (Fig. 4e). In this regard, three of 44 RPI1 regulatory genes are downregulated and five are upregulated in [SWI + ] cells (Fig. 4e left). The potential effectors can be Crz1 (downregulated repressor), Mig2, and Rpi1 self (upregulated activators) among these regulators. Similarly, among the 27 RDS1 regulators, two are upregulated and three are downregulated in [SWI + ] cells, and the potential regulators causing the upregulation can be Rpn4 (upregulated activator), Aft1, and Rds1 self (downregulated repressors) (Fig. 4f). The upregulation of Rds1 and Rpi1 may neutralize the effects of their aggregation to certain extent. These data provide one additional regulatory mechanism that may impact the [SWI + ]-transcriptome. ] and swi1∆, noticeable overlapping effects include enrichments in endopeptidase inhibitor activity, anchored plasma membrane components, and a subgroup of cell wall constituents (Fig. 5a-c).The enrichment for cell wall components is consistent with the lack of flocculin genes expression and our earlier reports on deficiency in multicellular features for both types of cells 31 . For the upregulated DEGs, both [SWI + ] and swi1∆ strains share an enrichment only in the carboxylic acid metabolic process. Other than the above common effects, most of the enrichments observed are specific to each of the four subgroups of our DEGs (Fig. 5). With regard to this notion, the most remarkable enrichment is in ribosome biosynthesis/assembly/transport, rRNA synthesis/ processing and other translation-related events, which are specifically downregulated by swi1∆ (Fig. 5). Interestingly, transcription of ribosome biogenesis genes is upregulated by [SWI + ] (Fig. 5). These data suggest that the pathway of protein translation is largely impaired in swi1∆ cells but may be promoted in prion cells. In contrast to the effects of swi1∆, the enriched pathways and functions for the downregulated DEGs in [SWI + ] cells are diverse, and are mainly related to oxidation-reduction, respiratory chain and other energy generating metabolisms, and antibiotic metabolism and response to external stimulus. For DEGs upregulated by swi1∆, specific enrichments were noticed for pathways and functions relevant to sexual reproduction/mating, metabolism of pyridoxine, amino acids and G-protein signaling (Fig. 5). Moreover, for DEGs upregulated by [SWI + ], specific enrichments were observed for RNA polymerase I-dependent ncRNA transcription and glutamine family amino acids biosynthesis (Fig. 5). When similar pair-wise comparisons of [SWI + ]-and swi1∆-DEGs were expanded to a broader range of processes (top 100), similar conclusions can be drawn (Fig. S5). Our enrichment analysis results partially overlap with those reported by Malovichko et al. with substantial differences as will be discussed later 44 .

GO term analysis for [SWI
We next analyzed whether [SWI + ] and swi1∆ differentially regulate the pathway of carbohydrate utilization. There are four carbohydrate-related processes, including carbohydrate catabolic process, cellular carbohydrate metabolic process, oligosaccharide metabolic process, and carbohydrate transmembrane transport, which are all enriched in DEGs repressed in [SWI + ] cells ( Fig. 5 and S5). However, carbohydrate metabolic process is the only pathway merely enriched in genes upregulated in swi1∆ cells. Thus, the carbon usage capacity may be impaired in [SWI + ] but not in swi1∆ cells. This seems to contradict our previous report that both swi1∆ and [SWI + ] cells are deficient in using non-glucose carbon sources 16 . To clarify this, further analyses were performed. We found that most identified DEGs that are involved in carbon usage are inhibited in both swi1∆ and [SWI + ] cells ( Fig. 5 and Table S1). In [SWI + ] cells, the inhibited genes encode proteins for maltose transporter (MAL31) and regulator (MAL33), hexose transporters (HXT5, HXT2, HXT13, HXT6, HXT7 (Table S1), suggesting a mechanistic difference in causing their deficiency in galactose usage in the two types of cells.

The response of [SWI + ] and swi1∆ cells to stress. Based on our GO term enrichment analysis predict-
ing that ribosomes and translation-linked biological processes might be dysfunctional in swi1∆ cells, we next carried out KEGG (Kyoto Encyclopedia of Genes and Gemomes) ribosome enrichment analysis for [SWI + ] and swi1∆ DEGs. As shown in Fig. 6a, an overrepresentation of KEGG ribosome genes was only seen for DEGs repressed by swi1∆, not for other three subgroups of DEGs. We speculated that swi1∆ cells may produce reduced amounts of protein. To test that, steady-state levels of proteins from the three yeast strains were compared after normalizing to number of cells. As shown in Fig. 6b (left), although [SWI + ] cells produced slightly more protein compared to wt, such a difference is not statistically significant. In comparison, swi1∆ cells show significantly lower amount of total proteins. In agreement with this observation, we found that swi1∆ cells indeed grow slower and have significantly a relative smaller cell size (Fig. 6b, right). We next examined whether swi1∆ and [SWI + ] cells respond to environmental stresses differently. Under stressful conditions, yeast may activate defense systems, concurrent with slowing growth as a protection mechanism. www.nature.com/scientificreports/ Stresses may activate a common gene expression program called Environmental Stress Response (ESR) 51,52 , which is comprised of ∼ 300 induced (iESR) transcripts involved in stress defense and ∼ 600 repressed (rESR) transcripts important for cell division. Our analyses demonstrated that the iESR and rESR genes are differently regulated in swi1∆ and [SWI + ] cells (Fig. 6c). Specifically, a significantly larger number of rESR genes are activated (9.0%) rather than repressed (0.0%), while more iESR genes are significantly repressed (47.1%) rather than activated (0.7%) in [SWI + ] cells. In other words, [SWI + ] cells seem to have activated rESR genes but repressed iESR genes in YPD, a rich medium where the growth condition is supposed to be not stressful, implicating that [SWI + ] cells may display a compromised ESR if cultivated in stressful conditions. Consistent with the characterization of slow growth, swi1∆ cells significantly reduce the transcription of rESR genes (25.3% repressed and 0.2% activated) but do not significantly alter the transcription of iESR genes (8.1% repressed and 4.3% activated) (Fig. 6c). These data implicate that different from [SWI + ] cells, swi1∆ cells would have partially activated ESR even when they grow in a rich medium like YPD. These data suggest that [SWI + ] and swi1∆ cells may be mechanistically distinct in responding to stresses. Based on the above analyses, we next performed experiments to test ESR of the swi1∆, [SWI + ], and wt strains. Cells were subjected to transient thermal stress, high osmolality, or drought treatment. Cell survivability was then determined on a rich medium, YPD after a given period of recovery time. As shown in Fig. 6d,e, both swi1∆ and [SWI + ] cells showed greater sensitivities to these stresses than that of wt cells.
We then examined cells' responses to various chemical stresses by growing them in liquid YPD medium supplemented with certain amounts of chemical stressors that were chosen based on literature from SGD (Fig. 7a). The fitness of cells was then determined by monitoring the growth curves (optical density at 600 nm-OD600) in a time window of 72 h. As shown in Fig. 7a, four growth patterns were discovered upon challenging with 34 stressors-no detectable differences, detrimental to both swi1∆ and [SWI + ], detrimental to swi1∆ only, and beneficial to [SWI + ] and/or swi1∆. We found that all three tested strains grew similarly for 7 chemical stressors (Congo Red, benomyl, boric acid, caffeine, hydrogen chloride pH 2.4, lysozyme, streptomycin), suggesting that no specific toxicities acted upon swi1∆ and [SWI + ] cells under these conditions. However, for 16 chemical molecules-cadmium dichloride, bleomycin B2, calcium dichloride, cycloheximide, ethanol, formaldehyde, iso-propanol, BAPTA, phosphorus, zinc dichloride, dimethyl sulfoxide, manganese sulfate, nickel dichloride, hydroxyurea, β-2-mercaptoethanol, acetic acid pH 4.3, most of these stressful conditions are detrimental to both swi1∆ and [SWI + ] cells in comparison to wild-type cells. Moreover, when challenged with 6 chemicals with the specified concentrations (paraquat dichloride, sodium m-arsenite, hydrogen peroxide, methanol, lytic enzyme, protease), a swi1∆-specific detriment was observed. Intriguingly, when supplemented with 5 additional chemical compounds, both swi1∆ and [SWI + ] cells grew better than wt cells though in some cases the difference is subtle (Fig. 7a bottom,b). Among the 5 compounds, 3 are alcohols-2-methyl-2-butanol, 1-butanol, and ally alcohol, while the other two are copper sulfate and 1, 4-dithiothreitol. Interestingly, ally alcohol is toxic to all three strains but [SWI + ] cells showed a better tolerance. Taken together, our data suggest that the deletion and prionization of Swi1 can have a wide range of distinct phenotypic effects on cells under stressful conditions, which can be either neutral, detrimental, or beneficial.

Discussion
The chromatin remodeling function by the ATP-dependent SWI/SNF complex is critical for various nuclear machineries to access DNA regions packed by histones and thereby essential for diverse cellular activities in the nucleus such as transcription, DNA replication, recombination and repair 22,53 . In this study, we have demonstrated that when Swi1, an important subunit of SWI/SNF, adopts a prion conformation to become [SWI + ], it exhibits a distinct transcriptional profile compared to that of wt and swi1∆ cells. Such a prion-specific transcriptomic signature is tightly associated with the altered transcription of genes encoding for TFs and Swi1 interacting proteins and the aggregation of numerous TFs. These prion-mediated modulations coordinate with substantial changes in various cellular pathways and functions. In contrast to a simple decline of protein translation-linked processes in swi1∆ cells, functional alterations caused by [SWI + ] are predicted to be largely distinctive to those caused by swi1∆. We show that [SWI + ] and swi1∆ cells exhibit diverse and complicated stress responses, indicating that the two strains may adopt distinct mechanisms in ERS.
In this study, we identified that ~ 15.7% of yeast genes are differentially expressed in swi1∆ cells (Fig. 1). Interestingly, an earlier study performed by Sudarsanam 44 . In contrast to their observation of a reduced expression of SUP45 in [SWI + ] and an elevated expression of ade1-14 in swi1∆ cells 44,55 , we found that SUP45 transcription is not altered and that ADE1 expression is rather mildly decreased in swi1∆ cells. With regard to PER33, ICS2, DLD3, and ENA1 that were shown to be uniquely upregulated in [SWI + ] cells in their study 44 , we found that the transcription is not altered in [SWI + ] cells for PER33, but reduced only in swi1∆ cells for ICS2, and increased for DLD3 or reduced for ENA1 in both swi1∆ and [SWI + ] cells (Table S1) (Table S1). Regarding this, chromosome I disomy may be a contributory factor for their observed change in HSP12 and HXT5 transcription in swi1∆ cells since it can lead to upregulation of OAF1 that encodes a TF activating the two genes 44 . In contrast to the Malovichko et al. study, we identified a large number of genes that are uniquely regulated either in swi1∆ or [SWI + ] cells -15 genes activated by [SWI + ] but inhibited by swi1∆, and 32 genes inhibited by [SWI + ] but activated by swi1∆ (Fig. 1e). As to GO term enrichment analyses ( Fig. 5 and 44   www.nature.com/scientificreports/ not noticed in our study. In addition, the upregulated pathways by both [SWI + ] and swi1∆ in this study were not found by Malovichko et al. Moreover, for DEGs upregulated by [SWI + ], we found a few enrichments while Malovichko et al. did not observe any ( Fig. 5 and 44 ). These discrepancies can be at least partially explained by different strains and cultivating conditions being used; Malovichko et al. used strains that may carry a mutated FLO8 gene and a functionally altered SUP35 gene 44 . Our strains carry no alterations at the SUP35 locus. In addition, chromosome I disomy of the swi1∆ strain in their study was not observed for our strains. Further, yeast cells were cultivated in a minimal medium containing galactose as sole carbon source in which both the swi1∆ and [SWI + ] strains would grow poorly in their study while our strains were grown in YPD, a standard rich medium. Therefore, the impact of different Swi1 states is condition-and genetic background-specific. As a global gene regulator, the SWI/SNF complex directly occupies promoter regions of ∼ 10% of yeast genes 24 , however, we identified a much larger number of DEGs from [SWI + ] and swi1∆ cells (Fig. 1a). In addition, the transcriptomic effects of [SWI + ] and swi1∆ can be distinct from different studies as discussed earlier. We believe secondary responses under specific cellular and environmental conditions besides the functional reduction of Swi1 may have specified such distinctions. For example, we found that the transcriptional and conformational alterations of TFs and Swi1 interacting proteins are important contributors of the modified transcriptome of [SWI + ] cells. Interestingly, we also found that the transcription of Swi1, Rds1, and Rpi1, which are aggregationprone in prion cells, is promoted (Fig. 4), suggesting a feedback regulation mediated by TFs. Since Swi1 activity reduction, aggregation of the TFs, and their combination cannot fully interpret the transcriptional alterations of their targets, we suspect that other alternative mechanisms may be involved, such as modifications on the integrity, functionality or specificity of the SWI/SNF complex and aggregation of other unknown proteins in [SWI + ] cells.
A recent study shows that there are 32 TF genes that are differently regulated in several mutants of SWI/SNF subunits (19 inhibited and 13 activated), and two of them, MET28 and MET32 contribute to Snf2 occupancy and transcriptional alterations following the loss of Snf5 and Swi3 26 . Unfortunately, the SWI1 deletion strain was not included in that study. Nevertheless, we have noticed that none of the 4 activated TF genes in swi1∆ cells from this study belong to the 32 TF genes, whereas among the 8 swi1∆-inhibited TF genes (Fig. S2), HMS1, TYE7, and HAP4 are similarly inhibited in those SWI/SNF subunit mutants 26 . Among the 7 [SWI + ]-activated TF genes, NRG2 and YPR015C are also activated in some of the mutants, however, RPI1 is inhibited and others are not changed in those mutants. NRG2 and RPI1 are also activated by swi1∆ in this study, suggesting that they are common inhibitory targets of these SWI/SNF subunits. Among the 15 [SWI + ]-inhibited TF genes, HMS1 is also inhibited by swi1∆ and by a few mutants of other SWI/SNF subunits, suggesting that an intact SWI/SNF complex is essential for its activation; however, instead of being inhibited, CAT3, MGA1, EDS1, and GAL3 are activated in those SWI/SNF mutants. These 4 TF genes may have specific contributions to the altered transcriptome of [SWI + ] cells. It was proposed that loss of some individual SWI/SNF subunits does not necessarily cause the collapse of the complex; instead, it can result in formation of altered sub-complexes that may significantly modulate Snf2 occupancy and thus alters the transcriptomic profile (e.g., number and regulatory pattern of DEGs) [24][25][26]56 . However, the effect of Swi1 loss on SWI/SNF architecture and Snf2 occupancy has not yet been tested. The lack of related data makes it less convincing to designate Swi1 as a separate module of the complex. Swi1 can interact with TFs individually 57 and can be targeted to the ARG1 promoter in the absence of Snf2 58 . In HeLa S3 cells, human Snf5 and Swi3 can be targeted to distinct DNA loci that are not bound by Snf2 59 . Under hypoxia, Snf11 and members of the Snf5/Swi3 regulatory module can migrate to the cytoplasm, whereas others may remain in the nucleus 60 . Thus, it is possible that Swi1 may act by itself or collaborate with some of the members of the SWI/ SNF complex in regulating the transcription of genes in alternative loci. In [SWI + ] cells, the aggregation of TFs in addition to Swi1 prion aggregation adds an extra layer of regulatory complexity. In addition to the three TFs we reported previously 31 , Mss11, Msn1, and Sap30, we have identified four additional TFs that are aggregated in [SWI + ] cells (Fig. 3). Of the TFs, Mot3 is particularly interesting as it is the protein determinant of the prion [MOT3 + ] 14 . Our result implicates that Swi1 and Mot3 may interact during the prionogenesis process and that they are likely to influence each other's de novo appearance and propagation. It remains unclear to what extent the SWI/SNF complex recruits the monomeric or oligomeric forms of Swi1 and how the structure and activity of the complex are altered in the context of [SWI + ]. It would be interesting to explore in future studies if chromatin interactions and histone marks shown as important factors of chromosome remodeling and gene expression 61 , are also modulated in [SWI + ] cells.
One of the most dramatic findings in this study is that genes for RNA processing, ribosome biogenesis and assembly, and other translation-related processes are severely inhibited in swi1∆ cells (Figs. 5, 6 and Table S1). Earlier studies showed that the loss of Snf5, Swi3, or Swi1 also has similar defects 26,44,62 . These data suggest that SWI/SNF is an important regulator of protein translation. Interestingly, most rESR genes are linked to translation, and stressed cells are programed to reduce cell division by inhibiting general protein production and to induce transcription of genes that are associated with cellular defense (mainly iESR genes) 52,63 . Our finding that swi1∆ cells inhibit rESR genes (Fig. 6c) suggests that swi1∆ cells may be partially under stress even in rich medium (YPD). Interestingly, [SWI + ] cells do not exhibit the defects in translation-related processes (Figs. 5, 6 and Table S1). On the contrary, ribosomal and rESR genes are activated in [SWI + ] cells, suggesting that cell growth is promoted in [SWI + ] cells. The fact that iESR genes are inhibited in [SWI + ] cells suggests the stress response is not triggered but rather somehow repressed in the prion cells. Taken together, swi1∆ and [SWI + ] cells may mechanistically adopt different strategies in ESR even though the outputs of ESR can be similar for the two strains (Figs. 6, 7). In addition, genes repressed by [SWI + ] have an overrepresentation of genes associated with the positive regulation of ion transport, response to water deprivation and salt stresses (data not shown), which may explain the sensitivity of [SWI + ] cells to thermal and osmotic stresses, and drought treatment (Fig. 6). This is consistent with earlier observations that Snf2, Snf5 and Taf14 are involved in the stress alleviation 64 www.nature.com/scientificreports/ than this, diverse responses of swi1∆ and [SWI + ] cells to a variety of chemical stresses highlights the complexity of ESR mediated by the Swi1 prion and its null mutant. Taken together, our study has revealed substantially differences of the transcriptomic profiles of [SWI + ] and swi1∆ cells, and important mechanisms contributing to the acquired transcriptional fingerprint and gain-offunction phenotypes of [SWI + ] cells. Our results may not only provide insight into mechanism governing the prion-mediated transcriptional regulation in yeast but also offer mechanistic clues for deciphering the pathogenesis of other aggregation-associated human diseases.

Methods
Strains media, oligo primers and plasmids. Wild-type non-prion ([swi -] or wt), [SWI + ] and swi1∆ strains used in this study were created and reported previously 31 . These isogenic Saccharomyces cerevisiae strains (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 flo8 FLO8-HIS3) are all S288C derivatives that are FLO8-repaired to enable the flocculin (FLO) gene expression and the only difference among them lies in the Swi1 state: swi1∆-the chromosomal SWI1 is deleted; [SWI + ]-wt genome but Swi1 is in prion conformation; and wt-wt genome and Swi1 is in non-prion conformation.
Standard yeast-cultivating rich media (YPD) or synthetic complete (SC) media with defined amino acid supply for selection were used in this study. For raffinose phenotype assay, 2% raffinose was the only carbon source in the SC media, which was supplemented with 0.5 µg/ml antimycin (Sigma-Aldrich, St. Louis, MO). Yeast strains were grown at 30 °C unless otherwise defined. Bacterial strains were grown in LB medium at 37 °C.

RNA-seq, read processing, and differentially expressed genes (DEG) calling. Overnight cultures
in YPD were diluted to ~ 10 5 cells/mL with fresh YPD. After 6 h of growth, cell concentrations of log-phase cultures were quantified and harvested. Total RNA was then extracted with RNeasy Mini Kit (Cat. No. 74104, QIAGEN) using protocols provided by the manufacturer. To avoid unnecessary perturbation of cellular metabolism that may alter gene expression profiles during enzymatic lysis, cells were lysed by bead-beating not by the recommended enzymatic lysis as described in the manufacturer protocol. Lysates were treated twice with DNase I to ensure a complete removal of genomic DNA. RNA QC and quantitation was performed using the Agilent Bio-analyzer 2100.
RNA-seq was conducted at University of Chicago Genomics Facility (http://genom ics.bsd.uchic ago.edu/) with three bio-repeats for each of the three strains. In brief, libraries were generated using the Illumina TruSEQ mRNA kit as protocolled by Illumina. Libraries were quantitated using the Agilent Bio-analyzer and sequenced on an Illumina HiSEQ2500 as protocolled by Illumina with a reading type of bidirectional 50-bp.
RNA-seq reads were trimmed to remove low quality trailing sequence using Trimmomatic 0.33 66 . Reads were then aligned to the sacCer3 assembly of the Saccharomyces cerevisiae genome with Tophat 2.1.0 67 and attributed to genes using HTSeq 0.6.1 68 . The depth of the sequencing was roughly estimated from the length of S. cerevisiae genome (G), the number of reads (N), and the average read length (L) as NxL/G. Differentially expressed genes were called using EdgeR 69 , with an adjusted p-value cutoff < 0.001 and |log2(FC)|> 1. We used the plotMDS function from EdgeR to assess the variance between samples, and two duplicates for each strain were used for final DEG calling.
Computational analyses. Comprehensive integrated biological information for the budding yeast Saccharomyces cerevisiae S288C was retrieved from the Saccharomyces Genome Database (SGD). The Venn diagram was generated using an online tool (http://bioin forma tics.psb.ugent .be/webto ols/Venn/). YeastMine 70 was used to generate gene lists and analyze the distribution and features of genes of interest lists. TF enrichment was performed using YEASTRACT 71,72 with a cut-off of p-value smaller than 0.001 based on DNA-binding plus expression evidence. The TF/target associations and regulatory networks were visualized with a force-directed layout. Aggregation-prone TFs were predicted with CamSol 49 and PrionW 50 , and 30 TFs were selected for aggregation assay when the intrinsic variant score is smaller than 1.5 with CamSol, or the pWALTZ score is larger than 73.5 and an N + Q content higher than 20% with PrionW. Metascape 73 was used for the Gene Ontology (GO) enrichment analysis. The enrichment cut-off used in this study was set to > 3 for Min Overlap, < 0.001 for p-value and > 1.5 for Min Enrichment. Gene Prioritization by Evidence Counting (GPEC, beta testing) was performed, otherwise a few analyses were also performed with a p-value of < 0.01. The visualization of enriched GO pathways was edited and optimized with Cytoscape 3.7.1 74 . To test if a gene list contains overrepresented TF genes, the ratio of TF genes in a user list was compared to the ratio of total TF genes (183 total TF genes defined by YEASTRACT) in the yeast genome (7185 genes), and Fisher's exact test was used to calculate the significance. Similarly, the overrepresentation of environmental stress response (ESR) genes including ~ 300 induced (rESR) and ~ 600 reduced (rESR) transcripts 51 as well as KEGG ribosome genes (https ://www.genom e.jp) in user gene lists were analyzed, and the significance was calculated with Fisher's exact test. In these significance tests, the cut-off of the p-value is < 0.05. Growth assays. Since swi1∆ and [SWI + ] cells exhibit abolished or reduced ability in using non-glucose carbon sources such as raffinose, glycerol and galactose 16 , the growth of swi1∆ and [SWI + ] strains were assayed on SC agar plates containing the indicated carbon source. After spreading cells onto plates with the indicated carbon source, images were taken after 3 days of incubation at 30 °C. As control, the swi1∆ strain was also trans-Scientific Reports | (2020) 10:21838 | https://doi.org/10.1038/s41598-020-77993-0 www.nature.com/scientificreports/ formed with p416SWI-SWI1, a single-copy plasmid expressing SWI1 from its endogenous promoter or an empty vector to verify the phenotypic complementation in similar growth assays using SC media without uracil. The effect on cell growth upon heat shock, freeze-thaw, drought, or high osmolarity stress was analyzed on semisolid agar plates. For heat shock treatment, properly diluted log-phase YPD cultures of wt, swi1∆ and [SWI + ] strains were incubated at 57 °C and spotted onto YPD plates at different time points. For freeze-thaw treatment, the same YPD cultures, part of which was used for heat-shock treatment, were frozen at − 80 °C and thawed at 42 °C for number of cycles as indicated before spotting onto YPD plates. In the drought assay, cells from the same cultures were pelleted by spinning down at 2500 rpm for 3 min, kept at room temperature for different days after removing the supernatant, and were spotted onto YPD. To examine cell viability in responding to osmotic stress, cells derived from log-phase YPD cultures were properly diluted into YPD liquid medium supplemented with 1 M NaCl, incubated at room temperature for a different number of days before spotting onto YPD plates. In these assays, the cell survivability was estimated after 3 days of growth. To examine cell growth in response to chemical stressors, overnight YPD cultures of wt, swi1∆, and [SWI + ] strains were counted and properly diluted into YPD liquid medium containing various amount of a chemical compound that is potentially toxic to yeast based on literature. The cell growth dynamics was then recorded by measuring the optical density at 600 nm (OD600nm) in a time course of 72 h.
Aggregation assay. The aggregation states of each examined protein in wt, swi1∆, and [SWI + ] strains were investigated by fluorescence microscopy 27,43 . In brief, the Swi1 state was assessed by examining the aggregation of Swi1-NQ-YFP expressed from a single-copy plasmid p416TEF-NQYFP that was controlled by the constitutive TEF1 promoter. The Rnq1 ([RNQ + ] determinant) status was verified by examining the aggregation status of Rnq1-GFP whose expression was governed by CUP1 promoter from a single-copy plasmid pCUP1-RNQ1GFP. For Swi1, cells from SC-ura were directly observed while the Rnq1 aggregation was examined after growing cells in SC-ura medium containing 25 µM CuSO 4 . For both prion proteins, an empty plasmid was included as a negative control. To examine the aggregation propensity of 30 TF-YFP proteins that were under control of GAL1 promoter from a single-copy plasmid, the wt and [SWI + ] strains were transformed, and the log-phase cultures (in SC-leu containing 2% sucrose) of their transformants were induced with different amounts of galactose (from 0.01 to 2%) with different induction times. Induction with 0.02% galactose for up to eight hours was finalized in the formal test. The aggregation frequencies were then observed with a fluorescence microscope and quantified.
Adhesive growth assay. Yeast multicellular phenotypes conferred by flocculin genes including adhesive growth on agar plates were confirmed based on methods described previously 31,75 . In brief, the tested strains were simply grown on YPD or SC selective plates, incubated for 3 days at 30 °C and then continued to incubate at room temperature for 3 days. Cells were then washed with water for different times with or without rubbing. Images were taken before and after washing.
Protein quantification in cell lysates.. This was done based on a published protocol 76 . Briefly, overnight cultures of wt, swi1∆ and [SWI + ] cells in liquid YPD were properly diluted into the same medium and harvested at log and stationary phases. Cell pellets were collected after centrifugation at 660 g for 3 min and washed twice with water. Cell densities were determined after cell counting. Cells were disrupted by glass beads for 6 × 1 min using a bead-beater in lysis buffer (50 mM sodium phosphate buffer, pH 7.5, 100 mM NaCl, 1 mM dithiothreitol, 2 mM PMSF, 2 μg/mL pepstatin A, 2 μg/mL leupeptin, 5 μg/mL aprotinin and 1 mM benzamidine-HCl). Lysates were centrifuged at 500 g for 5 min at 4 °C to remove cell debris. Supernatants were transferred to new tubes and the protein concentration was determined using Pierce 660 protein assay reagent (Thermo Scientific, ProD# 22,660) based on a protocol provided by the manufacturer. Different concentrations of BSA were used as control to set the standard of protein concentrations. Multiple tests were performed with at least three bio-repeats in each test. Protein concentrations were then normalized to cell numbers as outcomes.
Scientific Reports | (2020) 10:21838 | https://doi.org/10.1038/s41598-020-77993-0 www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.