Abstract

   Rho GTPases play critical roles in cell proliferation and cell death in
   many species. As in animal cells, cells of the budding yeast
   Saccharomyces cerevisiae undergo regulated cell death under various
   physiological conditions and upon exposure to external stress. The
   [42]Rho5 GTPase is necessary for oxidant-induced cell death, and cells
   expressing a constitutively active GTP-locked [43]Rho5 are
   hypersensitive to oxidants. Yet how [44]Rho5 regulates yeast cell death
   has been poorly understood. To identify genes that are involved in the
   [45]Rho5-mediated cell death program, we performed two complementary
   genome-wide screens: one screen for oxidant-resistant deletion mutants
   and another screen for [46]Rho5-associated proteins. Functional
   enrichment and interaction network analysis revealed enrichment for
   genes in pathways related to metabolism, transport, and plasma membrane
   organization. In particular, we find that [47]ATG21, which is known to
   be involved in the CVT (Cytoplasm-to-Vacuole Targeting) pathway and
   mitophagy, is necessary for cell death induced by oxidants. Cells
   lacking [48]Atg21 exhibit little cell death upon exposure to oxidants
   even when the GTP-locked [49]Rho5 is expressed. Moreover, [50]Atg21
   interacts with [51]Rho5 preferentially in its GTP-bound state,
   suggesting that [52]Atg21 is a downstream target of [53]Rho5 in
   oxidant-induced cell death. Given the high degree of conservation of
   Rho GTPases and autophagy from yeast to human, this study may provide
   insight into regulated cell death in eukaryotes in general.

   Keywords: yeast knockout strains, bimolecular fluorescence
   complementation, oxidative stress, regulated cell death, autophagy
     __________________________________________________________________

   Rho GTPases regulate diverse cellular processes in species ranging from
   yeast to humans. In addition to their functions in cytoskeleton
   organization in various cell types ([54]Hall 2012), Rho GTPases
   including Rac have been implicated in phagocyte killing and programmed
   cell death (PCD) in Caenorhabditis elegans ([55]Ravichandran and Lorenz
   2007). The evolutionally conserved role of Rac during the engulfment of
   apoptotic cells in mammals and C. elegans is related to their critical
   roles in cytoskeleton reorganization ([56]Ravichandran and Lorenz
   2007). Rac GTPases are also involved in activation of NADPH oxidases
   (NOX enzymes), which accept electrons from NADPH to produce the
   superoxide radical, in neutrophils and non-phagocytic cells ([57]Abo et
   al. 1991; [58]Pick 2014; [59]Werner 2004). Budding yeast has nine ORFs
   with sequence similarity to mammalian NADPH oxidases, and some of them
   are involved in the regulation of the actin cytoskeleton
   ([60]Rinnerthaler et al. 2012), but their link to Rho GTPase is not
   known.

   A large number of studies have shown that yeast cells undergo
   ‘regulated cell death (RCD)’ or PCD under various physiological
   conditions ([61]Carmona-Gutierrez et al. 2018; [62]Strich 2015).
   Regulation of cell death appears conserved in yeast, sharing some
   common regulators of cell death in metazoan and other multicellular
   systems, including the AAA-ATPase [63]Cdc48/VCP ([64]Madeo et al. 1997;
   [65]Braun and Zischka 2008; [66]Braun et al. 2006) and metacaspases
   ([67]Madeo et al. 2002). Yet the mechanisms by which yeast cell death
   is regulated are not well understood. We previously found that
   [68]Rho5, which is closely related to Rac GTPases in mammals, is
   necessary for oxidant-induced cell death in budding yeast ([69]Singh et
   al. 2008). Since [70]Rho5 interacts with [71]Trr1, thioredoxin
   reductase, specifically in its active GTP-bound state, we proposed that
   [72]Rho5 might downregulate the thioredoxin anti-oxidant system during
   cell death ([73]Singh et al. 2008). Other studies have suggested that
   [74]Rho5 downregulates the yeast cell wall integrity pathway
   ([75]Schmitz et al. 2002) and is involved in osmotic stress response
   ([76]Annan et al. 2008), although the underlying mechanisms are not
   clear. Consistent with these previous reports, cells lacking [77]DCK1
   and [78]LMO1, the homologs of mammalian DOCK180 and ELMO, which
   catalyze the nucleotide exchange of Rac1 or Cdc42 ([79]Brugnera et al.
   2002; [80]Cote and Vuori 2002), exhibit hyper-resistance to cell wall
   stress and hydrogen peroxide (H[2]O[2]) ([81]Schmitz et al. 2015).

   Since Rac is an important player during apoptotic cell death in other
   cell types, we asked whether a similar mechanism might be involved in
   [82]Rho5-mediated cell death in yeast. The [83]rho5^G12V mutant, which
   is believed to encode the GTP-locked [84]Rho5 in vivo, exhibits higher
   sensitivity to oxidants compared to cells lacking [85]TRR1 ([86]Singh
   et al. 2008), suggesting that [87]Rho5 has additional targets to
   promote cell death. We thus performed genome-wide screens to identify
   genes that are closely associated with [88]Rho5 and are likely involved
   in oxidant-induced cell death. Here, we report that several genes
   involved in vesicular traffic and organelle organization are important
   for oxidant-induced cell death. In particular, we found that [89]ATG21,
   which is known to be involved in the CVT pathway and mitophagy
   ([90]Strømhaug et al. 2004), is necessary for cell death mediated by
   [91]Rho5.

Materials And Methods

Plasmids, yeast strains, and growth conditions

   The Saccharomyces cerevisiae haploid knockout (YKO) strains (Thermo
   Scientific Open Biosystem) and wild-type (WT) BY4741 were used to
   screen for deletion mutants that were resistant to oxidants. A
   collection of VN (the N-terminal fragment of Venus, a yellow
   fluorescent protein)-tagged yeast strains ([92]Sung et al. 2013) was
   used to screen for [93]Rho5-binding proteins by bimolecular
   fluorescence complementation (BiFC) assays. All yeast strains and
   plasmids used in this study are listed in Supplemental Table S1 and S2,
   respectively, with a brief description. Standard methods of yeast
   genetics, DNA manipulation, and growth conditions were used
   ([94]Guthrie and Fink 1991). Yeast strains were grown in rich yeast
   medium YPD (yeast extract, peptone, dextrose) or synthetic complete
   (SC) containing 2% dextrose as a carbon source, unless stated
   otherwise.

Growth phenotype and treatment With H[2]O[2] or heat stress

   Sensitivity to H[2]O[2] was monitored by plating assays, as previously
   described ([95]Singh et al. 2008). Since the laboratory WT strains
   exhibited a varying degree of sensitivity to H[2]O[2] depending on the
   background ([96]Singh et al. 2008), 1 mM H[2]O[2] was used for all
   strains in BY4741 background, whereas 3∼4 mM H[2]O[2] was used for
   strains in HPY210 background to test oxidant-induced cell death. All
   strains were examined in comparison to WT in the isogenic strain
   background. Treatment with ramped heat stimulus was performed as
   previously described ([97]Teng et al. 2011): cells in fresh culture
   (OD[600] = 0.5∼0.6) were treated by increasing temperature in a PCR
   machine, first from 25° to 40° over 2 min and then 40° to 51° during
   the 10 min period. Cells were then kept at 51° for 10 min and plated on
   YPD. Cells grown for 1 day (postdiauxic) were treated by increasing
   temperature from 25° to 55° during the 15 min period and then kept at
   55° for 10 min before plating on YPD.

Cell viability assays

   Plasma membrane integrity was monitored by staining cells with
   propidium iodide (PI) immediately after incubation with 4 mM H[2]O[2]
   for 4 hr (or mock-treated), as previously described ([98]Kainz et al.
   2017). Single z-stack images of PI-stained cells were captured using a
   Nikon E800 microscope with a 40x objective lens and TRITC/TexasRed
   filter from Chroma Technology, Hamamatsu ORCA-2 CCD (Hamamatsu
   Photonics), and Slidebook software (Intelligent Imaging Innovations).
   The methylene blue reduction test was used to determine metabolically
   active cells, which convert methylene blue to colorless leucomethylene
   blue ([99]Bapat et al. 2006). After H[2]O[2] treatment (or
   mock-treated) as described above, cells were stained with 10 μM
   methylene blue (Sigma) and then observed by DIC (Differential
   interference contrast microscopy) using the Nikon E800 microscope with
   a 100x/1.30 NA oil-immersion objective lens. The same H[2]O[2]-treated
   and mock-treated cells were also plated on YPD plates (∼200 cells per
   plate) to determine the colony forming unit (CFU).

Genome-wide screen for H[2]O[2]-resistant mutants

   First, the YKO collection in 96-well plates was transformed with a
   multi-copy [100]rho5^G12V plasmid to generate a ‘sensitized’ strain
   collection, and the transformants were selected on SC-Ura plates. Each
   transformant as well as WT strain carrying the same plasmid organized
   in the 96-well plates (in duplicate sets) were treated with 1 mM
   H[2]O[2] for 3 hr or mock-treated, and then plated on SC-Ura plates
   using a 96-pin metal multi-blot replicator (‘Frogger’, V&P Scientific,
   Inc). After two days of incubation at 30°, deletion mutants that were
   more resistant to H[2]O[2] than WT were visibly identified. Those
   potential candidate strains were subjected to retesting under the same
   conditions except by growing them in individual liquid culture using
   SC-Ura medium. The authenticity of each deletion that was reproducibly
   resistant to H[2]O[2] were then tested by genomic PCR. Second,
   deletions of these candidate genes were generated in another strain
   background (HPY210) and tested for resistance to oxidants with 4 mM
   H[2]O[2] (see above).

Genome-wide screen by BiFC assays

   Briefly, a haploid strain HY1029, which expresses VC (the C-terminal
   fragment of Venus) fused to the N terminus of [101]Rho5, was mated with
   each VN fusion strain in 96-well plates, followed by selection of
   diploids on SC-Met-Lys, as previously described ([102]Sung et al.
   2013)(see Table S1). Diploid cells were grown to mid-logarithmic phase
   at 30° in SC medium, and initial screens were performed using 96-well
   glass-bottomed microplates (MGP096, Matrical Bioscience) and a Nikon
   Eclipse E1 microscope and a Plan Fluor 100×/1.30 NA oil immersion
   objective. Fluorescence images were obtained by a PhotoFluor LM-75
   light source (89 North Inc.).

Microscopy and image analysis

   Since the same split site (154/155) was used to generate both EYFP and
   Venus truncated forms for BiFC, candidate VN fusions from the initial
   screen were tested for nucleotide-specific interactions with [103]Rho5
   using YFP^C (the C-terminal fragment of YFP) fusions of WT and
   [104]Rho5 mutant proteins. Cells were grown in the appropriate
   synthetic medium overnight and then freshly sub-cultured for 3∼4 hr in
   the same medium prior to imaging. For images shown in Figure S2, 3
   z-stack (0.4 μm step) images were captured at room temperature (23∼25°)
   using a Nikon E800 microscope fitted with a 100x/1.30 NA oil-immersion
   objective lens and a YFP filter from Chroma Technology (see above).
   Out-of-focus signals were removed by deconvolution using Slidebook
   software (Intelligent Imaging Innovations). For images shown in
   [105]Figure 5A, slides were prepared similarly on an agarose slab, and
   3 z-stack (0.4 μm step) images were captured at 22° using a spinning
   disk confocal microscope (Ultra-VIEW VoX CSU-X1 system; Perkin
   Elmer-Cetus) equipped with a 100x /1.4 NA Plan Apochromat objective
   lens (Nikon); 440-, 488-, 515- and 561-nm solid-state lasers (Modular
   Laser System 2.0; Perkin Elmer-Cetus); and a back-thinned EM CCD
   (ImagEM C9100-13; Hamamatsu Photonics) on an inverted microscope (Ti-E;
   Nikon).

Figure 5.

   [106]Figure 5
   [107]Open in a new tab

   Rho5 interacts with Atg21 preferentially in its GTP-bound state. A.
   BiFC analyses of strains expressing Atg21-VN together with YFP^C fusion
   of WT or mutant Rho5. (a) Representative images (summed intensity
   projections) are shown for each strain as indicated. Size bar = 5 μm.
   (b) Quantification of YFP signals (mean ± SEM) from each BiFC analysis
   is plotted from 4 independent image sets. Each mark represents average
   fluorescent intensity from individual cells above threshold (n = 10 ∼25
   for each mark in case of WT and Rho5^G12V; n = 2∼3 for each mark in
   case of Rho5^K16N) (see Materials and Methods). Student’s t-tests were
   used and depicted with the following notation: ns (not significant) for
   P ≥ 0.05 and ***P < 0.001. B. (a) Yeast two-hybrid analyses of strains
   expressing a DBD fusion of WT or mutant Rho5 (or vector control)
   together with AD-Atg21 or vector control, as depicted. Three
   independent transformants of each plasmid combination were patched on
   SGal-Leu (also lacking Trp and His) to monitor interactions between
   Rho5 and Atg21 (right) and on SGal+Leu (lacking Trp and His) as a
   control (left). (b) Immunoblot showing LexA fusions to the WT and
   mutant Rho5 proteins (marked with an arrow) detected with polyclonal
   antibodies against LexA; an asterisk marks a non-specifically
   cross-reacting protein as a loading control. C. Cell survival of WT and
   atg21Δ cells (in BY4741 background) after treating with 1 mM H[2]O[2]
   for 4 hr or mock treatment. Student’s t-tests were used and depicted
   with the following notation: ns (not significant) for P ≥ 0.05; *P <
   0.05; and **P < 0.01. D. Fivefold serial dilutions of WT, atg21Δ, and
   atg18Δ (all in BY4741 background) carrying pRS-rho5^G12V plasmid were
   spotted after treatment with 1 mM H[2]O[2] for 4 hr or mock treatment.

   Image processing and analyses were performed using ImageJ (National
   Institutes of Health). Summed intensity projections of z stacks of
   representative images were used to generate [108]Figure 5Aa. To
   quantify BiFC signals, summed intensity projections were analyzed after
   background subtraction as follows: First, a fluorescence threshold was
   set above background that was estimated from a control strain (that
   expressed untagged [109]Rho5 together with [110]Atg21-VN). This
   threshold mainly selected fluorescent pixels on the vacuolar membrane.
   About 40∼62% of cells expressing YFP^C-[111]Rho5 or
   YFP^C-[112]Rho5^G12V had YFP signals above threshold (n = 86 for WT; n
   = 156 for [113]Rho5^G12V). In contrast, less than 10% of cells
   expressing YFP^C-[114]Rho5^K16N (n = 142) had YFP signal above
   threshold. To quantify BiFC signals in individual cells, all selected
   pixels above threshold in each cell were measured by drawing ROI around
   the cell boundary (based on the synchronized DIC image). The same
   threshold was also applied to all three strains that were captured and
   processed using the same conditions. Average integrated density of all
   ROIs (depicted with each mark on the graph) was calculated for each
   image set, and mean ± SEM from analyses of 4 independent image sets are
   shown in [115]Figure 5Ab.

Yeast two-hybrid assay

   A two-hybrid assay was carried out as previously described ([116]Gyuris
   et al. 1993). [117]Atg21 was expressed as an activation domain (AD)
   fusion protein using plasmid pJG4-5-[118]ATG21 (see Table S2). WT and
   mutant [119]Rho5 proteins were expressed as DNA-binding domain (DBD)
   fusions using plasmids pEG202-[120]RHO5, pEG202-[121]rho5^G12V, and
   pEG202-[122]rho5^K16N, all of which also carry the C328S substitution,
   as previously described ([123]Singh et al. 2008). The yeast strain
   EGY48 carrying the [124]LEU2 reporter gene was transformed with
   pJG4-5-[125]ATG21 and each of pEG202-[126]RHO5 plasmids (or empty
   vector controls), and several independent transformants from each
   transformation were plated on SC-Gal plates lacking Leu (and Trp and
   His to maintain the plasmids) to test the [127]Rho5-[128]Atg21
   interaction. To compare the levels of WT or mutant [129]Rho5 proteins
   fused to LexA, yeast extracts were prepared from the strain (EGY48)
   carrying each pEG plasmid and then were subjected to immunoblotting
   with rabbit polyclonal anti-LexA antibodies (EMD Millipore) and Alexa
   Fluor 680 goat anti-rabbit IgG (LI-COR Biosciences).

Gene Ontology (GO) term enrichment analysis

   Pathway enrichment analyses of the lists of genes derived from
   genome-wide studies were performed using the DAVID (Database for
   Annotation, Visualization and Integrated Discovery) bioinformatics
   resources ([130]Huang da et al. 2009a, [131]2009b). The GO for
   biological processes (BP), molecular functions (MF), cellular
   components (CC), and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathways that were significantly enriched are shown with respective p
   values (corrected for multiple comparison by Benjamini) in [132]Figure
   3C and Tables S3–S5.

Figure 3.

   [133]Figure 3
   [134]Open in a new tab

   A genome-wide screen for deletion mutants that are resistant to
   H[2]O[2]. A. A scheme of the screen for deletion mutants that are
   resistant to H[2]O[2] in a Rho5-dependent manner. A ‘sensitized’ YKO
   collection was created by transforming with pRS426-rho5^G12V and then
   subjected to H[2]O[2] treatment or mock-treatment. B. Fivefold serial
   dilutions of representative deletion mutants resistant to H[2]O[2]: (a)
   deletion mutants (BY4741 background) carrying pRS426-rho5^G12V, and (b)
   deletion mutants (HPY210 background) without the plasmid. C. Enriched
   GO terms from the ‘sensitized’ YKO screen.

Statistical analysis

   Data analyses and graph plotting were performed using Prism 6 (GraphPad
   Software). Mean and SEM (error bars) are provided in the bar graphs. A
   two-tailed student’s t-test was performed to determine statistical
   differences between two sets of data: ns (not significant) for P ≥
   0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

Data availability

   Strains and plasmids are available upon request. The VN fusion library
   is distributed by the Bioneer Corp ([135]www.bioneer.com). The authors
   affirm that all data necessary for confirming the conclusions of the
   article are present within the article, figures, and the supplemental
   files. Tables S1 and S2 contain lists of yeast strains and plasmids
   used in this study. Tables S3–S5 include all genomics data and GO term
   enrichment analyses by DAVID. Figures S1 shows spot assays of selected
   deletion mutants to test resistance to H[2]O[2]. Figure S2 shows BiFC
   analyses of candidate VN fusions with YFP^C fusions of WT or mutant
   [136]Rho5. Supplemental material available at Figshare:
   [137]https://doi.org/10.25387/g3.7294976.

Results And Discussion

Rho5 is necessary for cell death upon exposure to oxidants and heat stress

   While oxidant-induced cell death depends on [138]Rho5 in budding yeast
   ([139]Singh et al. 2008), it remained unclear how [140]Rho5 promotes
   cell death. Since yeast medium generally contains high levels of
   antioxidants, typically a higher concentration of H[2]O[2] is used to
   introduce oxidative stress in budding yeast than other cell types. In
   addition, the sensitivity to H[2]O[2] differs significantly depending
   on strain background ([141]Singh et al. 2008). To investigate regulated
   cell death and to avoid accidental cell death caused under harsh
   conditions, it was thus critical to establish a proper experimental
   condition for each strain background. To ensure the [142]Rho5-mediated
   cell death conditions, we characterized [143]rho5 mutants by staining
   with PI or methylene blue (MB) immediately after treating with H[2]O[2]
   and also determined CFU following the same treatment with H[2]O[2]. PI
   is highly charged and therefore normally cell impermeant, but it
   penetrates damaged membranes ([144]Stan-Lotter et al. 2006). Prior to
   H[2]O[2] treatment, less than 3% of WT, [145]rho5Δ, and [146]rho5^G12V
   cells were PI-positive (data not shown), and the percentage of
   PI-positive cells was slightly increased in the [147]rho5^G12V cells
   after treatment with H[2]O[2] ([148]Figure 1A). Similarly, there was
   only a minor difference in the percentage of MB-negative cells between
   WT and [149]rho5∆ ([150]Figure 1B), suggesting the majority of cells
   have metabolic activity immediately following the treatment with
   H[2]O[2]. Yet the CFU was significantly different among these strains:
   while over 80% of [151]rho5Δ cells were viable, less than 10% of
   [152]rho5^G12V cells and 30% of WT cells were viable upon treatment
   with 4 mM H[2]O[2] ([153]Figure 1C). Thus, the majority of these cells
   likely entered early apoptotic stage (PI^-) immediately upon exposure
   to H[2]O[2], and cell death likely occurred during subsequent
   incubation without further stress. We observed that a lower H[2]O[2]
   concentration (that caused cell death to a majority of WT cells in
   S288C background) had little effect on cell survival in the HPY210
   background (compare [154]Figures 1C & 5C).

Figure 1.

   [155]Figure 1
   [156]Open in a new tab

   Rho5 mediates regulated cell death triggered by H[2]O[2] or heat ramp.
   A. PI staining of each strain (HPY210 background) right after exposure
   to 3 mM H[2]O[2] for 4 hr. Average percentage of PI^+ cells are
   indicated from three sets of data (n = 100∼200 for each sample per
   test). Mock-treated cells showed low numbers of PI^+ cells (2∼3.5%) for
   all strains (not shown). B. MB staining of each strain (HPY210
   background) after exposure to H[2]O[2] treatment for 4 hr or
   mock-treated. Mean ± SEM (error bars) are shown from three sets of data
   (n = 100 for each sample per test). C. CFU was measured for each strain
   (HPY210 background) after H[2]O[2] treatment for 4 hr. Mean ± SEM
   (error bars) are shown from three sets of data (n = 100 for each sample
   per test) D. Cell survival was determined by a fivefold serial dilution
   of each strain (BY4741 background) grown in YPD after subjecting to
   heat ramp: (a) cells without treatment; (b) cells from fresh cultures
   (OD[600] = 0.5∼0.6) (b); and (c) cells grown 1d (postdiauxic).

   A previous report has shown that a ramped heat stimulus, rather than
   sudden heat shock, results in reproducible gene-specific survival
   phenotypes ([157]Teng et al. 2011). By following a similar procedure,
   we tested viability of [158]rho5Δ when subjected to a controlled heat
   ramp (see Materials and Methods). We found that the [159]rho5Δ mutant
   was more resistant to heat stress than WT during the exponential phase
   of growth and postdiauxic shift ([160]Figure 1D). These observations
   suggest that [161]Rho5 promotes apoptosis upon exposure to various
   stresses in addition to oxidants, consistent with previous studies
   ([162]Schmitz et al. 2002; [163]Singh et al. 2008; [164]Annan et al.
   2008; [165]Schmitz et al. 2015).

Cell death mediated by Rho5 may be independent of known apoptotic factors in
budding yeast

   A number of orthologs of mammalian genes involved in apoptosis have
   been implicated in RCD in budding yeast. These include [166]YCA1
   (metacaspase) ([167]Madeo et al. 2002), [168]AIF1 (an ortholog of
   mammalian Apoptosis-Inducing Factor) ([169]Wissing et al. 2004),
   [170]NUC1 (mitochondrial nuclease, an ortholog of mammalian endoG)
   ([171]Büttner et al. 2007), and [172]STE20, a PAK (p21-activated
   protein kinase) ([173]Ahn et al. 2005). We asked whether cell death
   mediated by [174]Rho5 involves these genes. If any of these gene
   products functions downstream of [175]Rho5 in the same oxidant-induced
   cell death pathway, we would expect that cells lacking such a gene
   might be resistant to oxidants even when [176]Rho5^G12V was expressed,
   which caused hypersensitivity to oxidants ([177]Singh et al. 2008). We
   thus tested deletion mutants of these apoptotic factors after
   transforming with a plasmid carrying [178]rho5^G12V or a vector
   control. Plate assays using serial dilutions of these transformants
   indicated that the expression of the GTP-locked [179]Rho5 caused cells
   to be severely sensitive to H[2]O[2] in these mutants ([180]Figure 2).
   In addition, all these deletion mutants (which were in the same genetic
   background as [181]rho5Δ) with the vector control were not as resistant
   to H[2]O[2] as [182]rho5Δ. These observations suggest that [183]Rho5
   promotes cell death independently from these previously known players
   in yeast cell death.

Figure 2.

   [184]Figure 2
   [185]Open in a new tab

   The GTP-locked Rho5 promotes cell death independently of known
   apoptotic factors in budding yeast. Each indicated deletion strain (in
   HPY210 background) carrying a multicopy RHO5 plasmid or vector control,
   as indicated, was treated with 4 mM H[2]O[2] for 4 hr (or mock
   treated), and then fivefold serial dilutions (starting from OD[600] =
   1) were plated on SC-Ura plates.

Genome-wide screen uncovers potential downstream mediators of Rho5 involved
in oxidant-induced cell death

   To identify additional players that function together with [186]Rho5 in
   oxidant-induced cell death, we carried out a genome-wide screen for
   mutants that failed to undergo cell death upon exposure to H[2]O[2]. To
   facilitate identification of genes that function downstream of
   [187]Rho5 (rather than those that function in oxidant-induced cell
   death independently of [188]Rho5), we generated a collection of
   ‘sensitized’ deletion mutants by transforming the [189]rho5^G12V
   plasmid into each deletion strain in the haploid YKO collection. A
   total of 4,506 deletion strains in the ordered arrays that carried the
   [190]rho5^G12V plasmid were successfully recovered and then screened
   for their resistance to H[2]O[2] ([191]Figure 3A).

   From the primary screen of the ‘sensitized’ strain collection, we
   initially identified 235 candidates that were more resistant to
   H[2]O[2]than the WT control strain carrying the same plasmid. We tested
   serial dilutions of each primary candidate without the [192]rho5^G12V
   plasmid and then confirmed correct ORF deletions of those that were
   reproducibly resistant to H[2]O[2]. We also tested the phenotype of
   additional candidates from those identified from a screen by BiFC
   assays (see below). Collectively, we identified 67 mutants that were
   more resistant to H[2]O[2] than WT. To confirm that the phenotype
   resulted from the expected deletion (and not any other mutation in the
   strain background), we generated a deletion of each candidate gene in
   another strain background and then tested its resistance to H[2]O[2].
   Many deletion mutants in this second test exhibited only minor or
   little resistance to H[2]O[2] (see [193]Figure 3Bb). The commonly used
   strain S288C (in which YKO was generated) has a mutated copy of
   [194]HAP1 ([195]Gaisne et al. 1999), which is involved in carbon
   catabolite activation of transcription and also gene expression in
   response to the levels of heme and oxygen ([196]Guarente et al. 1984).
   S288C also has a variant allele of [197]MIP1, which increases petite
   frequency ([198]Dimitrov et al. 2009). These factors might have
   contributed to hypersensitivity of the YKO strains to oxidants.
   Together from this secondary test, we identified 31 genes whose
   deletion confers resistance to H[2]O[2] (Figure S1; Table S3).

   To gain insight into the biological processes that were represented in
   the genes identified from our screen, we then performed pathway
   enrichment analysis using the DAVID bioinformatics resources
   ([199]Huang da et al. 2009a) (see Materials and Methods). GO terms and
   KEGG pathways that were significantly enriched could be grouped into
   main categories of processes associated with GTP binding, intracellular
   trafficking and the plasma membrane, and metabolism ([200]Figure 3C;
   Table S3). These results suggest that [201]Rho5 may integrate signals
   from various intracellular subcompartments to mediate oxidant-induced
   cell death.

Genome-wide screen by BiFC assays uncovers potential targets of Rho5

   In parallel to the screen of the yeast deletion library described
   above, we carried out another genome-wide screen by BiFC assays to
   identify proteins associated with [202]Rho5. We screened the diploid
   library in ordered arrays, which were generated by mating each VN
   strain in the VN fusion library ([203]Sung et al. 2013) with a strain
   expressing VC-[204]Rho5 ([205]Figure 4A). In addition, we tested those
   positives that were identified only from the screen of the deletion
   library (see above) by BiFC assays. Collectively, we identified 44
   genes, whose VN fusions exhibited positive BiFC signals ([206]Figure
   4B; Table S4), including [207]ATG21 and [208]CWP1, which were also
   identified from the screen for H[2]O[2]-resistant mutants. These
   positive candidates may associate closely with [209]Rho5, although they
   may be necessarily involved in oxidant-induced cell death.

Figure 4.

   [210]Figure 4
   [211]Open in a new tab

   A genome-wide screen for Rho5 interactome by BiFC assays. A. A scheme
   of the screen of VN fusion library by BiFC assays. B. Examples of
   positive candidates identified by the BiFC-based screen with VC-Rho5.
   C. Venn diagram depicting the number of candidate genes identified from
   two genome-wide screens. D. Venn diagram depicting functional groups
   enriched from both genome-wide screens.

   To uncover the cellular processes that [212]Rho5 may control via these
   potential target genes, we performed pathway enrichment analysis and
   found that the 44 genes identified from the BiFC-based screen were
   significantly enriched in pathways associated with metabolism, cell
   wall processes, and intracellular trafficking (Table S4). When we
   compared genes identified from both genome-wide screens, we found 19
   genes that were positive in both categories ([213]Figure 4C), with gene
   enrichment (with P < 0.05) in transport and organelle organization
   ([214]Figure 4D; Table S5). Our initial screen with VN-[215]Rho5,
   instead of VN-[216]Rho5^G12V, might have missed a specific [217]Rho5
   effector that might interact transiently with the GTP-bound [218]Rho5.
   However, BiFC assays facilitate detection of transient protein-protein
   interactions often by resulting in irreversible bimolecular fluorescent
   complex formation ([219]Miller et al. 2015), and thus a [220]Rho5
   effector could be easily trapped with [221]Rho5-GTP. Some of these
   candidates exhibited GTP-dependent interactions with [222]Rho5, while
   others appeared less specific to the GTP-bound state of [223]Rho5 (see
   below).

A subset of positives identified from both genome-wide screens may interact
with Rho5 preferentially in its GTP-bound state

   Since small GTPases often interact with their downstream targets in a
   GTP-dependent manner, we tested by BiFC analyses whether the 16
   candidates that were identified from both genome-wide screens associate
   with [224]Rho5 in a nucleotide-specific manner. By examining strains
   expressing each candidate VN fusion together with YFP^C-fusions of WT
   or mutant [225]Rho5, we identified seven VN fusions that exhibited
   stronger fluorescence signals with YFP^C-[226]Rho5^G12Vor
   YFP^C-[227]Rho5^WT compared to YFP^C-[228]Rho5^K16N, which is in either
   GDP-locked or nucleotide-empty state in vivo (see [229]Figures 5 & S2),
   as discussed below. Since YFP^C fusions of WT and mutant [230]Rho5
   proteins are expressed at about the same level ([231]Singh et al.
   2008), these observations suggest that these 7 candidates interact
   preferentially with [232]Rho5-GTP.

   Interestingly, a major group of genes that were identified from both
   genome-wide screens included eisosome components. Eisosomes are
   protein-based structures associated with specific lipid domains on the
   plasma membrane furrows, which are known as the MCC (membrane
   compartment containing [233]Can1) domains in yeast ([234]Walther et al.
   2006; [235]Karotki et al. 2011; [236]Douglas and Konopka 2014).
   [237]Lsp1, a core component of eisosome ([238]Walther et al. 2006),
   showed positive BiFC signals with WT [239]Rho5 and [240]Rho5^G12V but
   little signal with [241]Rho5^K16N. Other proteins closely associated
   with eisosomes, [242]Sur7 and [243]Nce102, also exhibited BiFC signals
   with [244]Rho5, although their interactions were not specific to its
   nucleotide-bound state (Figure S2). We found that deletion mutants of
   the eisosome components, [245]lsp1Δ, [246]nce102Δ, [247]sur7Δ, and
   [248]pil1Δ, were resistant to cell death, albeit moderately, when
   exposed to oxidants or when subjected to controlled heat ramp (Figure
   S1, A & B). These findings suggest that [249]Rho5-GTP might signal to
   eisosomes via [250]Lsp1 under the conditions triggering apoptotic cell
   death.

   Potential involvement of eisosome components in [251]Rho5-mediated cell
   death is surprising, because previous studies in yeast and other fungi
   have suggested that the components of the MCC/eisosomes may play
   multiple roles in resistance to stress ([252]Foderaro et al. 2017).
   Another previous study, however, also reported that deletion of
   [253]PIL1 and [254]LSP1 enhances heat stress resistance ([255]Zhang et
   al. 2004), consistent with our findings. Although their molecular
   functions are controversial, recent studies have revealed that eisosome
   proteins play a role in the control of signaling pathways, especially
   TORC2 activation ([256]Fröhlich et al. 2014; [257]Kabeche et al. 2014;
   [258]Berchtold et al. 2012). TORC2 serves as a stress sensor to induce
   various cellular responses including actin polarization, sphingolipid
   biosynthesis, and stress-related gene transcription ([259]Loewith and
   Hall 2011). Given the possibility of eisosomes as a link between
   different signaling pathways, it will be interesting to further
   investigate the involvement of eisosome proteins in [260]Rho5 signaling
   pathway.

   [261]Cwp1 is another potential player in [262]Rho5-mediated cell death
   pathway. Interestingly, [263]CWP1 was previously identified as one of
   the genes that were essential for cell death induced by the expression
   of human apoptosis initiator caspase-10 in budding yeast
   ([264]Lisa-Santamaría et al. 2012). The [265]Cwp1-[266]Rho5 BiFC
   signals appeared at endomembranes including the endoplasmic reticulum
   (ER) (Figure S2), consistent with the localization of [267]Cwp1-GFP and
   GFP-[268]Rho5 ([269]Tkach et al. 2012; [270]Huh et al. 2003; [271]Singh
   et al. 2008) but not at the cell wall or birth scar as reported
   ([272]Smits et al. 2006). However, the location of BiFC signals may not
   necessarily reflect the site of interaction of the two proteins because
   of the possibility of irreversible bimolecular fluorescent complex
   formation. Further studies are required to validate the subcellular
   location where these proteins interact with each other and to determine
   the functional significance of this interaction.

   Other candidates that exhibit GTP-dependent BiFC signals with [273]Rho5
   include [274]Msp1 and [275]Tom70 (Figure S2), which act in sorting and
   transport of mitochondrial proteins ([276]Nakai et al. 1993). Since the
   mitochondrion is an intracellular platform integrating cell death and
   energy production ([277]Eisenberg et al. 2007; [278]Balaban et al.
   2005), it will be interesting to investigate further to understand the
   physiological significance of their interactions with [279]Rho5.
   [280]Vph1 and [281]Smi1 (also known as [282]Knr4) displayed
   GTP-dependent BiFC signals with [283]Rho5 (Figure S2), although their
   deletions exhibited relatively mild resistance to H[2]O[2] (Figure S1).
   [284]Vph1 is a subunit of vacuolar ATPase V[0] domain, which is located
   on the vacuolar membrane ([285]Manolson et al. 1992). Since V-ATPase
   has been shown to be involved in activation of TORC1 in response to
   cytoplasmic pH, which in turn depends on glucose ([286]Dechant et al.
   2014), this result suggests a possible link between [287]Rho5 and TORC1
   signaling. [288]SMI1 encodes a protein localized to the bud neck with a
   proposed role in coordinating cell cycle progression with cell wall
   integrity ([289]Martin-Yken et al. 2003), but any functional link
   between [290]Smi1 and [291]Rho5 remains unknown.

   Unlike these candidates discussed above and [292]Atg21 (see below),
   several other candidates did not exhibit a preferential interaction
   with [293]Rho5-GTP (see Figure S2). While these proteins may possibly
   interact with a [294]Rho5 domain that does not undergo a drastic
   conformational change depending on its GDP- or GTP-bound state, further
   investigation is necessary to validate these interactions with
   [295]Rho5 in vivo. Nevertheless, it is interesting to find that genes
   involved in glucose metabolism are enriched in our screens. [296]Hxt1
   and [297]Hxt3 are low affinity glucose transporters ([298]Lewis and
   Bisson 1991; [299]Ozcan and Johnston 1999), and [300]Hxk2, which is a
   predominant hexokinase during yeast growth on glucose, is involved in
   regulation of expression of [301]HXT1 ([302]Ozcan and Johnston 1999;
   [303]Johnston 1999). A large number of studies have shown that glucose
   metabolism can regulate cell death in variety of cell types
   ([304]Gottlob et al. 2001; [305]Zhao et al. 2008a; [306]Zhao et al.
   2008b; [307]Rathmell et al. 2003; [308]Matsuura et al. 2016).
   Interestingly, a recent study suggests that [309]Rho5 relocates from
   the plasma membrane to mitochondria upon glucose starvation
   ([310]Schmitz et al. 2018). Although the functional significance of
   this observation is yet to be determined, it is tempting to speculate
   that there may be a crosstalk between [311]Rho5-mediated cell death
   pathway and glucose signaling.

Rho5 may promote cell death via Atg21 Under oxidative stress

   Among the candidates that might function in [312]Rho5-mediated cell
   death, we looked at [313]Atg21 more closely because [314]atg21Δ
   exhibited strong resistance to H[2]O[2] to a similar extent as
   [315]rho5Δ in two strain backgrounds used in this study and even when
   [316]Rho5^G12V was expressed (see [317]Figures 3Bb, [318]5C & [319]5D).
   BiFC assays with specific [320]Rho5 fusions suggested that [321]Atg21
   interacts preferentially with [322]Rho5-GTP. The BiFC signal was mainly
   observed on the vacuolar membrane with some puncta near and/or on the
   vacuolar membrane ([323]Figure 5Aa), consistent with the localization
   of [324]Atg21 ([325]Strømhaug et al. 2004) and a part of [326]Rho5
   localization ([327]Singh et al. 2008). Quantification of the BiFC
   signals indicated that there were strong fluorescence signals in cells
   expressing YFP^C-[328]Rho5^G12V and slightly weaker signals in cells
   expressing YFP^C-[329]Rho5^WT, whereas only minor background level of
   fluorescence was observed in cells expressing YFP^C-[330]Rho5^K16N
   ([331]Figure 5Ab).

   As an alternative way to test the GTP-dependent interaction between
   [332]Rho5 and [333]Atg21, we performed a yeast two-hybrid assay. An
   AD-[334]Atg21 fusion was expressed together with a DBD fusion of WT or
   mutant [335]Rho5 in a strain carrying the [336]LEU2 reporter. Growth of
   the cells expressing AD-[337]Atg21 specifically with
   DBD-[338]Rho5^G12V, but not with DBD-[339]Rho5^K16N, on a plate lacking
   Leu indicated that [340]Rho5-GTP interacts with [341]Atg21 ([342]Figure
   5Ba), despite the similar level of DBD fusions of WT and mutant
   [343]Rho5 proteins ([344]Figure 5Bb). The positive interaction between
   [345]Rho5^WT and [346]Atg21 was more evident by BiFC assays compared to
   the two-hybrid assays. This is likely because the bimolecular
   fluorescent complex of YFP^C-[347]Rho5 and [348]Atg21-VN might be more
   stable or irreversible compared to the interaction between
   DBD-[349]Rho5^WT and AD-[350]Atg21 in a two-hybrid assay.

   Consistent with our initial screen by spot assays, an [351]atg21Δ
   mutant exhibited higher cell survival upon exposure to H[2]O[2]
   ([352]Figure 5C). Because the WT [353]ATG21 strain undergoes little
   cell death in the absence of [354]Rho5 and vice versa, and because
   [355]atg21Δ cells exhibit little cell death even when [356]rho5^G12V is
   expressed (see [357]Figure 3Ba), it is likely that [358]Rho5 and
   [359]Atg21 function together to promote cell death upon exposure to
   oxidants. Moreover, the interaction between [360]Rho5-GTP and
   [361]Atg21 described above suggest that [362]Atg21 may function
   downstream of [363]Rho5. [364]Atg21 is known to be involved in the CVT
   pathway and autophagy of mitochondrion ([365]Kim and Klionsky 2000).
   Interestingly, a previous report suggests that cells deleted for
   [366]RHO5 or its exchange factor ([367]dck1 or [368]lmo1) exhibit
   reduced mitophagy ([369]Schmitz et al. 2015), supporting the idea that
   [370]Rho5 may be activated upon exposure to oxidants and promote cell
   death via [371]Atg21. [372]ATG18 and [373]ATG21 are paralogous genes
   that encode WD-repeat proteins, similar to human WIPI2 (WD repeat
   domain, phosphoinositide interacting 2) ([374]Dove et al. 2004;
   [375]Strømhaug et al. 2004). However, this role of [376]Atg21 in
   oxidant-induced cell death is unlikely to be shared with [377]Atg18,
   since [378]atg18Δ expressing the GTP-locked [379]Rho5 did not exhibit
   resistance to H[2]O[2] ([380]Figure 5D).

   The interconnection between apoptosis and autophagy is still poorly
   understood. Autophagy is generally known to be a cell survival
   mechanism, but recent studies suggest that autophagy-dependent cell
   death (ADCD) is a regulated cell death mechanism that requires
   components of the autophagy machinery instead of other cell death
   pathways. Although it is more likely that ADCD depends on components of
   the autophagic machinery rather than autophagic responses, increasing
   evidence suggests that these pathways are intricately connected
   ([381]Cooper 2018; [382]Bialik et al. 2018; [383]Carmona-Gutierrez et
   al. 2018). Previous studies have shown that the expression of human
   apoptosis initiator caspase-8 or caspase-10 is toxic to budding yeast
   by inducing a lethal phenotype with hallmarks of apoptosis and
   autophagy ([384]Lisa-Santamaría et al. 2009). Interestingly, this
   caspase-10-induced cell death in yeast requires genes involved in
   autophagy ([385]Lisa-Santamaría et al. 2012). While further
   investigation is required to fully understand how [386]Rho5-GTP
   promotes cell death via [387]Atg21, our findings suggest an interesting
   possibility that a Rho GTPase may utilize components of the autophagy
   machinery to promote cell death under oxidative stress.

   In summary, our genome-wide studies reported in this study suggest that
   [388]Rho5-mediated cell death signaling might impact cellular functions
   involving vesicle-mediated transport and the plasma membrane. These
   findings further substantiate the possibility that the oxidant-induced
   cell death process is complex and requires an interplay of a wide range
   of intracellular processes. It is tempting to speculate that common
   cellular components or pathway such as autophagy or eisosomes might be
   utilized either for cell survival or for death depending on upstream
   signaling molecules and/or the level of stress. Indeed, despite the
   similar structure of Rho GTPases, [389]Rho1, which is involved in the
   cell integrity pathway ([390]Levin 2011), is necessary for survival
   under oxidative stress ([391]Lee et al. 2011), whereas [392]Rho5 is
   necessary for cell death ([393]Singh et al. 2008; this study). To fully
   understand the functional significance of these interactions and the
   complex signal integration for survival and death will require further
   investigation.

Acknowledgments