https://orcid.org/0000-0001-5265-1970
Figures
Abstract
Clonal communities of single celled organisms, such as bacterial or fungal colonies and biofilms, are spatially structured, with subdomains of cells experiencing differing environmental conditions. In the development of such communities, cell specialization is not only important to respond and adapt to the local environment but has the potential to increase the fitness of the clonal community through division of labor. Here, we examine colony development in a yeast strain (F13) that produces colonies with a highly structured “ruffled” phenotype in the colony periphery and an unstructured “smooth” phenotype in the colony center. We demonstrate that in the F13 genetic background deletions of transcription factors can either increase (dig1D, sfl1D) or decrease (tec1D) the degree of colony structure. To investigate the development of colony structure, we carried out gene expression analysis on F13 and the three deletion strains using RNA-seq. Samples were taken early in colony growth (day2), which precedes ruffled phenotype development in F13, and from the peripheral and central regions of colonies later in development (day5), at which time these regions are structured and unstructured (respectively) in F13. We identify genes responding additively and non-additively to the genotype and spatiotemporal factors and cluster these genes into a number of different expression patterns. We identify clusters whose expression correlates closely with the degree of colony structure in each sample and include genes with known roles in the development of colony structure. Individual deletion of 26 genes sampled from different clusters identified 5 with strong effects on colony morphology (BUD8, CIS3, FLO11, MSB2 and SFG1), all of which eliminated or greatly reduced the structure of the F13 outer region.
Citation: Cromie GA, Tan Z, Hays M, Sirr A, Dudley AM (2024) Spatiotemporal patterns of gene expression during development of a complex colony morphology. PLoS ONE 19(12): e0311061. https://doi.org/10.1371/journal.pone.0311061
Editor: Hector Escriva, Laboratoire Arago, FRANCE
Received: September 11, 2024; Accepted: November 7, 2024; Published: December 5, 2024
Copyright: © 2024 Cromie et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The datasets generated during and/or analyzed in the current study are available in the Gene Expression Omnibus (GEO) under accession GSE274952.
Funding: This work was funded by a grant (P50 GM076547) from the National Institutes of Health (https://www.nih.gov/) and a strategic partnership between the Institute for Systems Biology (https://isbscience.org/) and the University of Luxembourg (https://www.uni.lu/en/). Z.T. was funded by the Agency for Science, Technology, and Research, Singapore (https://www.a-star.edu.sg/). The funders played no role in the study design, data collection and analysis, decision to publish, or preparation of this manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Differentiation, the process by which cells specialize into distinct types, which can then be organized into larger-scale functional structures, is traditionally thought of as a phenomenon specific to multicellular organisms. In such organisms, differentiation increases fitness through structural and metabolic division of labor. Although the concept of differentiation is not meaningful in the context of isolated single-celled organisms, it is relevant to clonal communities of such organisms, such as bacterial or fungal colonies and biofilms [1]. In these spatially-structured communities, cells in different regions may experience significantly different environments, including gas and nutrient availability, concentrations of waste and signaling products, exterior environmental challenges, and pH [2]. In these situations, cell specialization may not only benefit the individual that needs to respond and adapt to the local environment, but by structuring the community into cooperating domains of specialized cells, may improve the fitness of the community as a whole [3].
Some wild isolates of the budding yeast Saccharomyces cerevisiae produce highly structured “ruffled” colonies, rather than the unstructured “smooth” colony morphology typical of most laboratory strains. These structured colonies exhibit many of the features of biofilms, including production of an extracellular matrix, increased adherence, and localized expression of drug efflux pumps [4–6]. Often, these colonies are unstructured in the early stages of colony growth but become structured as the colony grows and ages. As part of this process, domains of specialized cells develop within ruffled colonies, with cells at the colony base forming pseudohyphae that invade the agar medium and anchor the colony, cells in peripheral regions upregulating drug efflux pumps and surface cells entering stationary phase, where they become more resistant to environmental insult [7]. Similarly, distinct gene expression profiles have been observed for the invasive cells within the agar, compared to the rest of the colony [8]. The molecular and regulatory pathways underlying ruffled colony formation overlap to a large extent with a group of related traits including mat formation and filamentous growth [9].
In a previous study [10], we identified six genes which when overexpressed repress the ruffled colony morphology of a strain (F45) derived from a cross between a sake-brewing strain and an Ethiopian white tecc strain. We then further tested the role of these genes in colony morphology by deleting them in another strain (F13) that derived from the same cross between the sake and white tecc parents. Because F13 colonies display a partially structured phenotype with ruffled morphology in the colony periphery and an unstructured (smooth) phenotype in the colony center, these gene deletions might alter the ruffled or smooth morphologies in the different regions of the colonies. Consistent with the overexpression results, deletion of all six genes promoted development of the ruffled morphology in the center of the F13 colonies, while maintaining the ruffled phenotype of the periphery. The six identified genes included the transcription factors DIG1 and SFL1, that operate in the filamentation MAPK cascade and the PKA/cAMP pathways respectively and are known to have roles repressing ruffled colony formation and related phenotypes [11–13].
Although a number of biological pathways have been shown to regulate colony morphology [9,14], the processes that underly the generation of structured versus smooth morphology as a colony develops are still poorly understood. Here, we explore that question by assessing gene expression in F13 colonies, both before and after development of the structured outer region, and from both the outer ruffled and inner smooth regions of older colonies. We then use deletions of SFL1 and DIG1 to produce fully ruffled F13 colonies and demonstrate that deletion of the transcription factor TEC1 results in completely smooth F13 colonies. The effect on gene expression of these genetic perturbations that increase and decrease the degree of colony morphology was then assessed. Our results identify spatiotemporal patterns of gene expression that reflect the development of a complex colony morphology.
Materials and methods
Yeast strains and media
Unless noted, standard media and methods were used for growth and genetic manipulation of yeast [15] The strains of S. cerevisiae used in this study are listed in Table 1.
RNA preparation and sequencing
After 2 days of growth on YPD (2% glucose) plates at 30° C, whole colonies, arrayed in a “checkerboard” pattern [16], were harvested by scraping cells off the surface of the agar plate. To obtain sufficient quantities of RNA, 3–5 colonies were pooled for each sample, with three biological replicate samples taken of each strain/genotype. For the wild-type F13 strain, after 5 days of growth the smooth interior of the colony and the structured outside region were isolated separately, with inside and outside samples from 3–5 colonies pooled as above. For the deletion derivatives of F13, which form either fully smooth (tec1Δ) or fully ruffled (sfl1Δ, dig1Δ) colonies, inside and outside samples were taken to match the proportions taken from the F13 colonies.
Following extraction by hot acid phenol [17], total RNA from the pooled colonies was quantified by Bioanalyzer (Agilent). 5 μg of total RNA for each sample was then processed using the Tru-Seq stranded mRNA kit (Illumina) following manufacturer instructions. Individual sequencing libraries were pooled and analyzed by paired-end, 51-nucleotide read sequencing in one lane of an Illumina HiSeq 2000.
Read-pair alignment
Read-pair alignment was carried out against the S288c reference (R64-1-1), with the FASTA and GFF files extended (S1 and S2 Files) to include ncRNAs and genes present in F45, a strain produced by the same cross that generated F13, but absent in S288c, as described previously [18]. Alignment was carried out using Bowtie2 (version 2.1.0) [19] with the parameters [-N 1 -I 50 -X 450 -p 6—reorder -x -S] and allowing 1 mismatch per read.
For each strain, read alignments were then converted to gene counts using featureCounts (version 1.4.0) in the Subread package [20], with the parameters [-a -o -t gene–g ID–s 2 -T 1 -p -P -d 50 -D 450]. Reads were not filtered based on mapping quality, and thus we have been cautious in our interpretation of counts of genes that have paralogs with similar sequences, or which contain large regions of low sequence complexity. Read sequences are available from the Gene Expression Omnibus under accession GSE274952. Gene count tables are provided in S3 File.
Identification of genes responding to genotype or colony age/region
Analysis of gene expression was carried out using the edgeR [v. 3.6.8] [21] package for R [22] based on the tables of raw counts produced by featureCounts (S3 File). Library sizes were first normalized using calcNormFactors (applying Trimmed Mean of M-values). Then the data was filtered to include only ORFs present in the S288c reference genome (genes with systematic names beginning with “Y”) and the counts for each gene in each library were converted to log2 counts per million reads, using a prior count of 20 to reduce the variance associated with low-expression genes. This produced a table with 4 genotypes, measured in triplicate, across 3 times/regions (day-2, day-5-inside and day-5-outside) giving a total of 36 normalized libraries (S3 File). Two factor ANOVA models were then fit to this table with the first factor representing genotype (4 levels: tec1Δ, wt, sfl1Δ, dig1Δ) and the second factor representing the spatiotemporal variables (3 levels: day-2, day-5-inside and day-5-outside). Model fitting was performed twice, with the first ANOVA model allowing only an additive effect of the two factors (type II ANOVA) while the second ANOVA model included an interaction term between them (type III ANOVA). Multiple hypothesis correction was carried out for the p-value of the individual main and interaction terms in ANOVA separately using the Holm method [23]. We also tested each additive and interaction model against a null model (single mean) using ANOVA, and again multiple hypothesis corrected the results for each of the two models individually. For comparing the relative number of genes responding to the genotype and spatiotemporal factors, we used an adjusted p-value cutoff of 0.01 for each of the two terms in the additive ANOVA. To empirically confirm the stringency of our approach, we carried out one hundred permutations of the 36 sample measurements for each gene randomizing their sample label (i.e. the unique combinations of genotype spatiotemporal). We calculated the additive and interaction ANOVA p-values for each of these permutations and used these to calculate that our adjusted p-value cutoffs (p<0.01 versus null) for the additive and interaction ANOVAs correspond to empirical false discovery rates of 1.0 x 10−05 and 3.7 x 10−05, respectively.
We then identified genes with highly variable expression, having a minimum log2 expression variance of 0.2 across the whole dataset. Next, genes with an adjusted interaction term p-value of 0.01 or less from the interaction ANOVA were set aside as representing potential non-linear interactions between the two factors. From these, a final set of genes showing a strong non-additive response to both factors were then identified as those genes with a minimum r-squared of 0.9 in the interaction model. Among the remaining genes, those responding strongly and significantly to the individual factors were then identified from the first, additive ANOVA using a p-value cutoff of 0.01 (versus a null model) and a minimum r-squared cutoff of 0.7 for the full additive model, and a p-value cutoff of 0.01 for the factor of interest (all p-values were multiple hypothesis corrected, as described above).
Characterization of gene expression patterns
For examination of patterns of gene expression responding to genotype, to the spatiotemporal factor, or additively to both, the individual effect of each factor was isolated by normalizing for any effect of the other factor. This was done by setting the mean log expression of each gene within each level of the second factor to zero. After this normalization, the expression profiles of genes responding strongly and significantly to each factor underwent hierarchical clustering by gene. This was done using the pheatmap command of edgeR, with the expression of each gene centered and scaled to mean = 0 and variance = 1 and using the “complete linkage” clustering method.
For genes responding to the genotype factor, the three highest level clusters were extracted directly from the hierarchical clustering results. For genes responding to the spatiotemporal factor, eight major classes of gene expression were identified from visual inspection of the initial hierarchical clustering. These patterns were encoded as follows (2 = high expression, 1 = intermediate expression, 0 = low expression) with each value applying to a single level of the spatiotemporal factor (day-2, day-5-inside, day-5-outside; each containing three replicates of the four genotypes): C1 = (2,0,0), C2 = (0,2,2), C3 = (2,2,0), C4 = (0,0,2), C5 = (2,1,0), C6 = (0,1,2), C7 = (0,2,0), C8 = (2,0,2). Each gene was then assigned to the class whose template pattern showed the highest correlation with the expression pattern of the gene. Expression heatmaps for all individual classes were produced using the pheatmap command, from the pheatmap package [24] in R.
Because the dataset includes expression measurements taken after deleting these genes, care must be taken in interpreting the expression patterns of TEC1, SFL1 and DIG1.
Functional enrichment of gene lists
Functional enrichment of S. cerevisiae gene lists was performed using g:Profiler [25] with a multiple hypothesis corrected (g:SCS) threshold of p<0.01.
Results
Strain F13 is a model for detecting both increases and decreases in the degree of colony structure
The haploid budding yeast strain F13, which was derived from a cross between a sake-brewing strain and an Ethiopian white tecc strain [26], displays an unusual, highly heritable, pattern of colony development on rich solid medium. Initially, the colony is unstructured, similar to common laboratory strains, but after three days of growth develops a clearly-defined, highly-structured outer ring, with a smooth, unstructured center (Fig 1). This morphology is unusual because most budding yeast genotypes that we have examined form either fully structured or fully smooth colonies. Colony morphology is both genetically and environmentally regulated, with strains reproducibly structured or smooth on specific media, but often structured on some media conditions and smooth on others [14]. A number of environment-sensing signaling pathways are known to contribute to this “switch” [14], with strain genotype then determining sensitivity to these signals.
The unusual colony morphology of late-stage F13 colonies, having both smooth and structured zones, suggested that F13 might allow the detection of both perturbations that increase and perturbations that decrease the degree of complex colony morphology. We previously demonstrated [10] that genetic perturbations expected to increase colony structure cause F13 to form fully structured colonies, with no smooth center at late time points (Fig 1). Specifically, we deleted genes encoding two transcription factors, Dig1 and Sfl1, that operate in the filamentation MAPK cascade and the PKA/cAMP pathways respectively, that have been previously shown to inhibit complex colony morphologies [11–13]. F13 colonies of these genotypes have already begun to develop weak colony structure after two days of growth (day-2), when F13 colonies are still fully smooth (Fig 1). Here, we additionally tested the ability of F13 to respond to genetic perturbations that decrease colony structure, by deleting TEC1, a gene that also encodes a member of the filamentation MAPK cascade, but which promotes complex colony morphology [13]. Consistent with the known role of Tec1 as a positive regulator of the MAPK signaling pathway, deletion of TEC1 produced colonies that remained fully smooth throughout their development (Fig 1). Therefore, it appears that strain F13 is indeed a model allowing detection both of perturbations increasing, and perturbations decreasing, colony structure.
Patterns of gene expression in F13
To explore the patterns of gene expression associated with the regional development of colony morphology over time in strain F13 we carried out RNA-seq on complete F13 colonies from day-2 of growth (unstructured) and on the outer (structured) and inner (unstructured) regions of 5-day-old F13 colonies. In addition, we carried out RNA-seq on day-2 and day-5 colonies of strains harboring deletions of DIG1, SFL1 or TEC1. Although colonies with these deletion genotypes were either fully smooth or fully structured at day 5, we also isolated comparable portions of the inner and outer regions of these colonies (Materials and methods). This allowed comparison of the transcriptional profiles from three biological replicates of entire day-2 colonies as well as the outside and inside of each day 5 colony for all four genotypes. Using a single genetic background (F13) to compare gene expression in smooth versus structured colonies or colony-regions eliminates the effects of genome-wide sequence variation on gene expression that could potentially confound comparisons between genetically diverged fully-smooth versus fully-structured strains.
To identify the major patterns of gene expression relating to genotype, colony age and colony region, we carried out a two-way additive ANOVA on the gene expression data (normalized log2 counts per million reads) (Materials and methods). The first factor represented the genotype and had 4 levels: tec1Δ, wild-type F13 (WT), sfl1Δ and dig1Δ. The second factor captured the spatiotemporal variables and had 3 levels: day-2 (whole colony), day-5 (outside) and day-5 (inside). We identified 2,978 genes that showed a significant response to the two factors under an additive model (type II Anova p<0.01, after multiple hypothesis correction). Using permutation testing (Materials and methods), we determined that our chosen significance threshold corresponds to an empirical false discovery rate of 1.0 x 10−5. Among the set of genes, examination of the p-values associated with the two factors indicated that many more genes (2,772 vs 344) showed a significant (p<0.01, after multiple hypothesis correction) change in expression in response to the spatiotemporal factor than to the genotype factor (Fig 2). A total of 221 genes showed a significant additive response to both factors (see below).
In an additive model, a smaller number of genes show a significant (p<0.01, multiple hypothesis corrected) response to the genotype factor (A) than to the spatiotemporal factor (B).
To isolate the major genotype-dependent and spatiotemporal patterns of gene expression we normalized our expression data for the effect of one factor and carried out hierarchical clustering to identify major patterns of gene expression associated with the remaining factor in isolation (for genes significantly responding to that factor, Materials and methods). We focused on genes with the strongest and clearest additive response to the two factors, i.e. genes having a high degree of variation in expression within the experiment (variance of normalized log counts > 0.2) that was well explained by the additive model (r2>0.7). A small number of genes whose expression was better explained by a non-additive interaction between the two factors (Materials and methods) were excluded at this point for later analysis (see below).
Interestingly, although deletion of TEC1, SFL1 or DIG1 was used to produce fully smooth or fully structured colonies in this study, the expression of these genes varies little across our samples, with the variance of normalized log counts falling below our threshold of 0.2 in all cases (once the matching knockout strain is excluded). These three genes encode signaling proteins whose activity is strongly regulated by phosphorylation, emphasizing that our study is focused on gene expression and not on regulatory processes, such as protein phosphorylation, that are independent of gene expression.
Genotype-dependent patterns of gene expression
After normalizing for the spatiotemporal factor and carrying out hierarchical clustering (Materials and methods), three major patterns of gene expression were visible, among genes with statistically significant responses to the genotype factor (Fig 3).
Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low). Heatmap genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, black font indicates unstructured samples, magenta font indicates structured samples. Genotype clusters (G1-G3) indicated.
The first pair of clusters (G1 and G2) identified genes with mirror patterns of gene expression in the order tec1Δ, WT, sfl1Δ, dig1Δ, with dig1Δ showing the highest expression in cluster G1 and the lowest expression in cluster G2 (Fig 4). The order of these genotypes corresponds to the associated degree of colony morphology seen at day 5, with tec1Δ fully smooth, F13 WT intermediate and sfl1Δ and dig1Δ fully ruffled, i.e. the expression level of these genes correlates (positively in cluster G1 and negatively in cluster G2) with the degree of colony structure associated with each genotype. Genes in cluster G1 are significantly enriched (p<0.01) for a number of GO terms, including “fungal-type cell wall” (GO:0009277; p = 2.71e-04), “extracellular region” (GO:0005576; p = 7.78e-04) and “regulation of establishment or maintenance of cell polarity” (GO:0032878; p = 5.67e-04), and for targets of the transcription factor Tec1 (TF:M01810_0; p = 3.38e-07) (S1 Table). Genes in cluster G2 are significantly enriched for several GO terms including “extracellular region” (GO:0005576; p = 1.78e-11), “fungal-type cell wall” (GO:0009277; p = 1.72e-09), “cell wall organization” (GO:0071555; p = 1.42e-05), “hydrolase activity, hydrolyzing O-glycosyl compounds” (GO:0004553; p = 7.59e-05) and “side of membrane” (GO:0098552; p = 2.51e-04) and for targets of the transcription factor Matalpha2-Mcm1 (TF:M08191; p = 1.58e-04) (S2 Table).
Gene expression patterns (after normalization for spatiotemporal factor) within each of the 3 genotype clusters displayed as heatmaps (left) and expression traces (right) of all genes within each cluster (grey lines). In the expression traces, the mean expression of the genes in the cluster is highlighted in red line with circles reflecting genotype (orange = tec1Δ, grey = wt, green = sfl1Δ, blue = dig1Δ). Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured).
The final cluster (G3) identified a group of only 6 genes (BSC1, PRM7, MGA1, AQY2, YLL053C, FIT2) with highest expression in sfl1Δ and lowest expression in dig1Δ. These genes are significantly enriched for several GO terms, including “water channel activity” (GO:0015250; 5.62e-04) and water transport (GO:0006833; 2.63e-03) (S3 Table).
Spatiotemporal patterns of gene expression
After normalizing for genotype and carrying out hierarchical clustering (Materials and methods), 4 major patterns of gene expression were visible, among genes with statistically significant responses to the spatiotemporal factor. These four major patterns could be grouped into two sets of “mirror” pairs. The first set consisted of one group of genes with higher expression in the day-5-outside sample and lower expression in the other two samples, and one group of genes with the reverse pattern, having lowest expression in the day-5-outside sample (Fig 5). Both of these groups were small (clusters C7 and C8, below).
Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1, (black font = unstructured, magenta font = structured).
The second set of mirrored expression patterns consisted of one large group of genes with higher expression at day-2 and lower expression in the day-5-inside sample, and another large group showing the reverse pattern (day-2 low, day-5-inside high) (Fig 5). Within this second set, a continuum of expression levels was observed in the day-5-outside sample, therefore we further split these two sets of genes based on whether the day-5-outside expression level was intermediate between the day-2 and day-5-inside samples, or whether it more closely matched either the day-2 or day-5-inside expression level (Materials and methods). This gave us three pairs of mirror clusters (C1-C6, Figs 6 and 7).
Gene expression patterns (after normalization for genotype) within each of the first four spatiotemporal clusters displayed as heatmaps (left) and expression traces (right) of all genes within each cluster (grey lines). In the expression traces, the mean expression of the genes in the cluster is highlighted in red line with circles reflecting genotype (orange = tec1Δ, grey = WT F13, green = sfl1Δ, blue = dig1Δ). Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured).
Gene expression patterns (after normalization for genotype) within each of the second four spatiotemporal clusters displayed as heatmaps (left) and expression traces (right) of all genes within each cluster (grey lines). In the expression traces, the mean expression of the genes in the cluster is highlighted in red line with circles reflecting genotype (orange = tec1Δ, grey = WT F13 green = sfl1Δ, blue = dig1Δ). Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1, (black font = unstructured, magenta font = structured).
The first pair of these “mirror” clusters (C1 and C2) consisted of expression patterns that correspond to colony age/size, namely genes with differential expression between day-2 and day-5 that exhibited little difference in expression between the outside and inside regions of the day-5 colonies. Genes in cluster C1 were repressed at day-5, and genes in cluster C2 were induced (Fig 6). Genes in cluster C1 are significantly enriched for a number of GO terms that include “extracellular region” (GO:0005576; p = 4.23e-06), “fungal-type cell wall” (GO:0009277; p = 1.25e-05) and “side of membrane” (GO:0098552; p = 7.91e-05 (S4 Table), while genes in cluster C2 are significantly enriched for several GO terms, including “cellular component assembly involved in morphogenesis” (GO:0010927; p = 6.06e-05), “fungal-type cell wall” (GO:0009277; p = 1.86e-03), “prospore membrane” (GO:0005628; p = 6.33–03) and “glyoxalase III activity” (GO:0019172; p = 9.42e-03) (S5 Table).
The second pair of clusters (C3 and C4) reflect expression changes specific to the inside of the day-5 colonies, with genes exhibiting similar expression levels between the day-2 and the day-5-outside samples, but different levels of expression in the day-5-inside sample. In cluster C3 the genes were repressed in the day-5-inside sample and in cluster C4 they were induced (Fig 6). No significant GO or KEGG term enrichment was seen for genes in cluster C3 (S6 Table). Genes in cluster C4 were significantly enriched for a number of GO terms including “monocarboxylic acid metabolic process” (GO:0032787; p = 2.61e-07), “cellular lipid metabolic process” (GO:0044255; p = 2.59e-06) “C-acyltransferase activity” (GO:0016408; p = 1.00e-03), “ammonium transmembrane transporter activity” (GO:0008519; p = 1.00e-03) and “fatty acid synthase complex” (GO:0005835; p = 3.48e-03) (S7 Table).
The third pair of clusters (C5 and C6) consisted of genes for which levels of expression in the day-5-outside sample was intermediate between that of the day-2 and day-5-inside samples. In cluster C5, expression was lowest in the day-5-inside sample while in cluster C6 expression was highest in that sample (Fig 7). Genes in cluster C5 were enriched for GO terms that include “cytoskeleton” (GO:0005856; p = 2.09e-12), “amino acid biosynthetic process” (GO:0008652; p = 3.42e-12), “cellular bud” (GO:0005933; p = 2.96e-11) and “heterocyclic compound binding” (GO:1901363; p = 2.80e-06) and the KEGG term “Biosynthesis of amino acids” (KEGG:01230; p = 4.54e-15) (S8 Table). Genes in cluster C6 were enriched for GO terms including “sorbitol transport” (GO:0015795; p = 2.37e-05), “monocarboxylic acid metabolic process” (GO:0032787; p = 7.41e-05) and “peroxisome” (GO:0005777; p = 3.28e-04) (S9 Table).
The final pair of clusters (C7 and C8) displayed expression patterns that were specific to the outside region of the day-5 colonies. These clusters contain a small number of genes with similar levels of expression in day-2 and day-5-inside samples, but that differed from expression in the day-5-outside sample. In cluster C7 (S10 Table), expression was high in the day-5-outside sample while in cluster C8 (S11 Table) expression was low in that sample (Fig 7). No significant GO or KEGG term enrichment was seen for genes in either cluster.
Genes responding additively to both genotype and spatiotemporal factors
Our analysis so far examined genes responding either to a single factor (genotype or spatiotemporal) or additively to both factors, allowing the effect of each factor to be studied in isolation and genes to be clustered into a discrete set of factor-specific expression patterns. Next, for genes responding to both factors, we examined how the genotype and spatiotemporal clusters interacted with one another, i.e. were genes in specific genotype clusters over-represented in a subset of spatiotemporal clusters? The number of genes responding additively to both factors was much larger than expected by chance, with 616 genes responding to the spatiotemporal factor, 129 genes responding to genotype, and 103 genes responding to both (Fisher’s Exact Test: one tailed p = 4.73e-84; 6575 genes total). That is, genes responding to the genotype factor are essentially a subset of those responding to the spatiotemporal factor. For the shared set of genes, the distribution of counts (Table 2) across the clusters of the two factors was clearly non-independent (Fisher’s Exact Test: two tailed p = 2.05e-08), indicating a relationship between the spatiotemporal and genotype expression patterns.
We next looked in more detail at the relationship between the spatiotemporal and genotype clusters, for genes responding to both factors. The expression pattern of genes in cluster G1 correlates positively with degree of fluffiness associated with each genotype. Genes in this cluster were over-represented in spatiotemporal cluster C2 (low expression in day-2 and high expression in both day-5 samples) and under-represented in cluster C1 (high expression in day-2 and low expression in both day-5 samples), compared to expectation under independence. In fact, out of all 85 genes in cluster G1 (regardless of any response to the spatiotemporal factor) a total of 39 (46%) were also placed in cluster C2. Taken together, these data suggest that there is a significant overlap between genes whose expression correlates positively with more structured genotypes and genes whose expression is higher in day-5, when ruffled morphology is most strongly developed, than in day-2, when it is weak or has not yet developed.
The opposite pattern was seen for genes in cluster G2 (gene expression correlates negatively with degree of fluffiness of genotype) whose expression also responded to the spatiotemporal factor. For these genes, an under-representation was seen in cluster C2 (low expression in day-2 and high expression in both day-5 samples) and an over-representation was seen in cluster C1 (high expression in day-2 and low expression in both day-5 samples), compared to the expectation under independence. That is, there is a strong overlap between genes whose expression correlates negatively with more structured genotypes and genes whose expression is higher in day-2 than in day-5.
Taken together, the overlap patterns seen with clusters G1, G2, C1 and C2 suggest the existence of a reciprocal relationship between expression in the structured vs smooth genotypes and expression in day-5 versus day-2. These patterns suggest that these two sets of genes (G1+C2 and G2+C1) might include genes important for the development of ruffled versus smooth colony morphologies.
The 39 genes assigned to both clusters G1 and C2 did not show any significant GO term enrichment (S12 Table) while the G2+C1 group (S13 Table) was significantly enriched for several related GO terms, including “extracellular region” (GO:0005576; p = 1.04e-04) as well as “fungal-type cell wall” (GO:0009277; p = 1.75e-04), and “side of membrane” (GO:0098552; p = 1.65e-03).
Genes responding non-additively to both genotype and spatiotemporal factors
We next examined the behavior of genes whose expression showed a significant non additive response to the genotype and spatiotemporal factors (Materials and methods). These genes were excluded from the preceding analyses as the effect of the spatiotemporal and genotype factors cannot be meaningfully studied in isolation when a non additive interaction exists. As a group (S14 Table), this set of genes show strong enrichment for the GO terms “cell periphery” (GO:0071944; p = 1.64e-10) and “extracellular region” (GO:0005576; p = 1.14e-08). Hierarchical clustering of these genes identified several distinct expression patterns, which we extracted as 9 clusters (Fig 8 and S11 Table).
Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured). Clusters I1 and I9 (asterisk) indicated.
One of these clusters (cluster I9), with eight members, included several genes that are known to have important roles in the development of complex colony morphologies. These genes are FLO11, encoding a cell wall flocculin believed to play a critical mechanistic role in pseudohyphal and filamentous growth and in the development of structured colonies [13,27,28], PHD1, encoding a transcriptional enhancer that promotes pseudohyphal growth (via mechanisms such as regulating FLO11 expression level) [29,30], and MSB2, encoding a cell wall mucin protein that promotes filamentous growth via the “filamentous” MAPK pathway [31]. These three genes either encode cell wall components or regulate cell wall processes and several other members of this cluster also have cell wall roles, with the paralogs SVS1 and SRL1 encoding cell wall proteins [32] and WSC2 encoding a signal transducer in the stress-activated PKC1-MPK1 signaling pathway which helps maintain cell wall integrity [33]. The final two genes are SIM1, encoding a SUN family protein that may participate in DNA replication [34], and TOS7, encoding a protein with likely roles in secretion and cell wall organization [35]. The genes in cluster I9 are enriched for several GO terms including “cell surface” (GO:0009986; p = 7.51e-05), “cell periphery” (GO:0071944; p = 4.35e-04) and “site of polarized growth” (GO:0030427; p = 5.12e-04) and the KEGG pathway “MAPK signaling pathway–yeast” (KEG:04011; p = 1.72e-03) (S15 Table).
The pattern of gene expression in cluster I9 in the day-5 samples correlates closely and positively with the degree of ruffled colony morphology in those samples (Fig 9). Expression levels are low in the tec1Δ day-5-outside sample and the tec1Δ and WT day-5-inside samples, which are all smooth. Conversely, expression is higher in the WT, sfl1Δ and dig1Δ day-5-outside samples, and the sfl1Δ and dig1Δ day-5-inside samples, all of which are ruffled. Notably, gene expression in the day-5 outside samples differs between tec1Δ (smooth) and F13 WT (ruffled) while expression is similar between these genotypes in the day-5 inside samples (both smooth). Expression in the day-2 samples correlates positively with the degree of fluffiness associated with each genotype at day-5, increasing in the order tec1Δ, F13 WT, sfl1Δ and dig1Δ, similar to cluster G1 in the analysis of the additive ANOVA.
Gene expression patterns within the spatiotemporal/genotype interaction clusters I1 and I9 displayed as heatmaps (left) and expression traces (right) of all genes within each cluster (grey lines). In the expression traces, the mean expression of the genes in the cluster is highlighted in red line with circles reflecting genotype (orange = tec1Δ, grey = WT F13, green = sfl1Δ, blue = dig1Δ). Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured).
Cluster I1 shows a pattern of gene expression that is essentially the inverse of cluster I9 across the day-5 genotypes and colony regions (Fig 9). However, expression in the day-2 samples varies little across the genotypes. Cluster I1 (S16 Table) consists of 8 genes (GDH3, MUP3, SYG1, SNZ1, FDH1, SSU1, OPT2 and YHR137C-A), showing no enrichment for any GO or KEGG pathways.
Changes in the expression of genes encoding extracellular proteins in response to the genotype and spatiotemporal factors
Several of the gene expression clusters that we identified in our analysis showed enrichment for genes encoding extracellular and cell wall proteins. This included the set of all genes showing a non-additive interaction between the genotype and spatiotemporal factors. In addition, several of the paired clusters that show inverse patterns of gene expression (e.g. C1 versus C2 and G1 versus G2) showed enrichment for extracellular genes in both of the clusters, despite their opposite patterns of expression. To look into the behavior of extracellular genes more directly we carried out cluster analysis of all genes with the GO terms (GO:0005576 “extracellular region”, GO:0009277 “fungal-type cell wall”, GO:0009986 “cell surface”, GO:0031505 “fungal-type cell wall organization”) and whose expression responded significantly to the additive model using the genotype and spatiotemporal factors described earlier or showed a significant non-additive response to the two factors (Materials and methods).
Based on this cluster analysis, it appears that the extracellular genes can be classified as responding mostly to the spatiotemporal factor, or mostly to the genotype factor (Fig 10 and S17 Table). Among the genes responding mainly to the spatiotemporal factor, two “reflected” patterns were observed with genes either showing highest expression at day-2 and lowest expression in the day-5-inside sample (Cluster E1), or vice versa (Cluster E2) (Fig 11). Genes encoding glycosylated cell surface proteins were observed in both clusters (E1: DAN1, FLO5, FLO9, TIR1, TIR2, TIR3, TIR4; E2: CWP2, CWP1, FIT2, PAU24, PHO5, PIR3, SED1) (S17 Table). In contrast, cluster E2 contained several spore wall genes (GAS4, SPO11, SPS22) and the two genes encoding a-factor (MFA1, MFA2) while cluster E1 contained genes encoding several enzymes involved in cell wall modification including glucanases (EGT2, EXG2, SUN4) and a chitin transglycosylase (UTR2) (S17 Table).
Color reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured).
Gene expression patterns within the extracellular gene expression clusters E1-5 displayed as heatmaps (left) and expression traces (right) of all genes within each cluster (grey lines). In the expression traces, the mean expression of the genes in the cluster is highlighted in red line with circles reflecting genotype (orange = tec1Δ, grey = WT F13, green = sfl1Δ, blue = dig1Δ). Color in heatmaps reflects log2 normalized expression for each gene (red high, blue low) and genotypes are indicated as T = tec1Δ, F = WT F13, S = sfl1Δ, D = dig1Δ, (black font = unstructured, magenta font = structured).
The genes responding mostly to the genotype factor fell into three main clusters. Two of these clusters showed patterns “reflected” for their behavior across the 4 genotypes in the day-5 samples, with expression higher in the tec1Δ and WT samples in cluster E4 and higher in the sfl1Δ and dig1Δ samples in cluster E5 (Fig 11). That is, the expression of these genes in the day-5 samples correlates positively or negatively with the degree of colony morphology seen in each genotype. Like clusters E1 and E2, genes encoding cell surface glycoproteins were found in both clusters E4 and E5, although these were more common in E5 (E4: DAN4, HPF1; E5: CIS3, FLO10, FLO11, MSB2, PIR1, SRL1) (S17 Table). Cluster E4 also contained both of the genes encoding alpha-factor (MF(ALPHA)1 and MF(ALPHA)2), SAG1 and AGA1 encoding alpha-agglutinin and the anchorage subunit of a-agglutinin, respectively, and several genes involved in cell wall remodeling during bud separation (CTS1, DSE2, DSE4). The third cluster mainly responding to genotype, E3, showed an expression pattern similar to E5 (Fig 11). This cluster consisted of a small number of cell wall genes (CWP1, CWP2, SED1, PAU24, NFG1), with 4/5 of these genes encoding known glycoproteins (the exception, NFG1, encodes a negative regulator of the filamentous growth pathway [36] (S17 Table).
Previously, using a full-structured strain, F45, isolated from the same cross as F13, we identified six genes whose overexpression caused colony morphology to transition from structured to smooth [10]. Expression analysis then identified sets of genes consistently upregulated or downregulated in the smooth overexpression strains relative to the initial, structured F45 strain. Interestingly, all of the genes from cluster E4 (higher expression in smooth day 5 regions) in this study (except for the two that encode alpha-factor) were also upregulated in the (smooth) F45 overexpression strains. In contrast, none of the genes in cluster E5 (lower higher expression in smooth day 5 regions) were consistently overexpressed in these strains, while 4/16 were repressed.
Looking more closely at clusters E4 and E5, it can be seen that the genes in each cluster vary in how closely their expression behavior at day-2 matches that at day-5. In each cluster, some genes show the same strong pattern of differential expression across genotypes at day-2 as they do at day-5, while other genes show a weaker version of this pattern and the remaining genes show little variation across genotypes at day-2 (Fig 11). These results are consistent with differential patterns of gene expression across genotypes that initiate at different timepoints in the development of the colony.
Effect on colony morphology of deleting genes from different transcriptional clusters
While characterizing gene expression patterns can yield valuable insights into biological processes, differentially expressed genes do not necessarily influence a phenotype of interest. Our analyses identified several transcriptional clusters characterized by common expression patterns in response to the genotype and spatiotemporal context of the cells. Several of these clusters appear to have meaningful correlations with colony morphology, both positive and negative. For example clusters G1 and I9 demonstrate gene expression that correlates positively with the degree of colony structure associated with each genotype while clusters G2 and I1 demonstrate a negative correlation.
To assess the effect of deleting genes with a variety of responses to the spatiotemporal and genetic factors, we deleted 26 genes in different clusters, with an emphasis on clusters that show correlations (positive or negative) with colony morphology. The genes that we chose to delete, and their cluster assignments are listed in Table 3. As expected from their known roles in development of complex colony morphologies [13,27,28], deletions of FLO11 or MSB2, from cluster I9, produced colonies that remained fully smooth at day-5 (Fig 12). Deletion of other I9 genes tested (PHD1, SIM1, SRL1 and SVS1) had little effect on F13 colony morphology, retaining a central smooth and outer structured zone at day-5. Among the remaining genes, deletion of SFG1 (clusters C3+G1) and CIS3 (clusters C5+G1) produced fully smooth colonies, while deletion of BUD8 (clusters C3+G1) produced colonies with a smooth center and notably reduced structure in the outer region of the colony. The remaining deletions had little effect on F13 colony morphology (Fig 12).
Discussion
Our study identified a number of coherent gene clusters in which expression varied by colony age and region and in the presence of genetic perturbations increasing or decreasing the degree of colony morphology. The major expression patterns associated with the genotype factor showed a striking positive or negative correlation with the degree of colony morphology achieved by colonies of each genotype on day-5. The major expression patterns associated with the spatiotemporal factor showed a strong difference between the day-2 and day-5-inside samples with the day-5-outside sample either resembling one of the other two, or displaying an intermediate expression level. Interestingly, some strong overlaps were seen between the genotype and spatiotemporal clusters, with an over-representation of genes that are more highly expressed in the structured genotypes and are also more highly expressed at day-5, when colonies can become fully ruffled, than at day-2, before colony structure has fully developed. A similar overlap was seen between the complementary clusters with highest expression in the unstructured genotypes and having highest expression at day-2.
Surprisingly, the genes positively correlated with the ruffled phenotype (by genotype and colony age/region) and those negatively correlated were enriched for many of the same GO terms, with over-representation of cell wall / extracellular genes in both directions. This suggests that extensive cell wall composition changes accompany changes in colony morphology, with numerous extracellular genes up- and down-regulated. When we looked specifically at the expression pattern of extracellular genes this pattern was confirmed, with some genes most highly expressed in the most structured genotypes and others most highly expressed in the smooth genotypes. For many of these genes, the molecular function of the proteins they encode is poorly understood consistent with them having roles in processes and pathways that are critical to complex colony morphology, but which have yet to be experimentally characterized. This is likely related to the fact that common laboratory strains of budding yeast, in which most experimental work has historically been performed, do not form structured colonies. For example, the reference strain S288c has acquired a mutation inactivating one of the major transcriptional promoters of complex colony morphologies [37]. Non-laboratory strains of budding yeast capable of producing complex colony morphologies, such as F13, are therefore attractive models in which to characterize these pathways. Our clustering results identify subsets of genes which may participate in common pathways, helping to inform future mechanistic studies.
Looking in more detail at the extracellular genes, among those responding to the spatiotemporal factor, we observed two “reflected” patterns, with genes either showing highest expression at day-2 or in the day-5-inside sample. Although the genes in each of these two clusters showed consistent expression behavior in the day-5 samples, the behavior of the genes at day-2 varied within each cluster. For some genes, the pattern of differential expression across genotypes that was seen in the day-5 samples was also seen at day-2, whereas for other genes this pattern was weaker or absent. This suggests that the time at which the pattern of differential expression first develops varies between genes. By day-5 all of these genes show differential expression across genotypes, largely correlating with the degree of colony structure observed at that time, but some genes are already strongly differentially expressed by day-2, when the tec1Δ and WT colonies are fully smooth and the sfl1Δ and dig1Δ colonies have only begun to show weak signs of colony structure. The genes showing early differential expression may be important for specifying the development of colony structure at later time points, i.e. they may be part of a “developmental program” for colony structure. In contrast, genes showing differential expression only after the full development of colony morphology may be physical effectors of the phenotype, or genes responding to different “environmental” conditions within ruffled versus smooth colonies. These results provide an additional level of information that may help inform the design of future mechanistic studies.
Deletion of a set of genes sampled from different clusters only identified five with strong effects on colony morphology (BUD8, CIS3, FLO11, MSB2 and SFG1), all of which greatly reduced or eliminated the structure of the F13 outer region. These five genes have diverse functional annotations. FLO11 and MSB2 encode cell surface glycoproteins with known roles in promoting colony structure [13,27,28], and CIS3 encodes a cell wall glycoprotein whose deletion causes loss of colony morphology in strain F45, related to the F13 strain used here [18]. SFG1 encodes a transcription factor known to promote pseudohyphal and invasive growth [36,38]. BUD8 is involved in bud site selection [39]. While we deleted a relatively small subset of genes displaying differential expression in our study, we did not observe any genes whose deletion caused F13 to become fully ruffled. This is consistent with the hypothesis that the complex structure observed in the outer ring of F13 may represent a specialized developmental pathway that elaborates on the “baseline” developmental program occurring in the smooth interior region. In this case, deletion of any of the genes promoting or effecting the specialized pathway would then cause loss of structure in the outer ring, resulting in fully smooth colonies. Under this model there are no “smooth-specific” developmental genes and only deletions of negative regulators of the ruffled developmental program, such as SFL1 or DIG1, would produce fully structured colonies.
The expression changes observed in our study were dominated by those associated with the spatiotemporal factor. The major clusters associated with this factor consistently showed a strong difference between the day-2 and day-5-inside samples, but were distinguished by the behavior of the day-5-outside sample. In some clusters the day-5 outside sample behaved like either the day-2 or day-5-inside samples, while in others it showed an expression pattern intermediate between the other two, i.e. genes showing unique expression in the day-5 outside samples were rare. Instead, the patterns that we do observe are consistent with gene expression in response to a buildup of signaling molecules or metabolic waste products within the colony over time, or conversely depletion of nutrients within the colony and adjacent regions of agar. Gene expression responding to these signals would place the day-2 sample and the day-5-inside sample at opposite extremes, with the day-5-outside sample showing intermediate expression, with its position relative to the other two depending on the steepness of concentration gradient(s) associated with the chemical(s) controlling gene expression, and any thresholds that exist for response to those gradients. Similarly, the oldest cells in the day-5 colony are likely to be in the inside region and the youngest in the outside region [40], so that average cell age could also vary from day-2 (youngest) to the day-5-inside sample (oldest) with the day-5-outside sample intermediate between the other two.
Several previous studies have identified patterns of spatial cell differentiation within yeast colonies. In addition to identifying differentiation within structured colonies [7,8], work from Palkova and colleagues has also identified patterns of vertical and horizontal cell differentiation within old smooth colonies. Such colonies display stratification into clearly demarcated upper (U) and lower (L) cell populations, with distinctive gene expression patterns [41], while programmed cell death is confined to the central region of old colonies [40]. The sharp demarcation between the U and L populations is similar to the clearly spatial defined subpopulation of sporulating cells observed by Honigberg and colleagues within diploid yeast colonies [42].
One outstanding question raised by our work and previous studies is the precise nature of the signals that specify spatial (and temporal) differentiation within yeast colonies. The sporulating region within diploid colonies has been shown to reflect an overlap between subpopulations expressing the IME1 and IME2 activators of sporulation and to respond to alkaline pH signaling through the Rim101/PacC pathway [42]. Similarly, the localization of programmed cell death to the inner regions of old colonies has been shown to be dependent on ammonia signaling [40]. However, the precise nature of the full set of environmental and cell-cell signals specifying patterns of yeast colony differentiation have yet to be fully elucidated in any system. The work presented here extends previous studies on yeast colony development to examine the interaction between genetic and spatiotemporal factors in the development of a highly structured colony morphology. Our results suggest a complex interplay between genetic perturbations and their effect on spatially defined subpopulations within a biofilm-model.
Supporting information
S1 File. S288c reference FASTA sequence, extended to include genes present in F45 and not S288c.
https://doi.org/10.1371/journal.pone.0311061.s001
(TXT)
S2 File. S288c reference GFF file, extended to include non-coding RNAs (ncRNAs) and genes present in F45 and not S288c.
https://doi.org/10.1371/journal.pone.0311061.s002
(TXT)
S3 File. Raw read counts and normalized log2 read counts per gene per library.
https://doi.org/10.1371/journal.pone.0311061.s003
(XLSX)
S4 File. R script used for data processing and analysis.
https://doi.org/10.1371/journal.pone.0311061.s004
(TXT)
S1 Table. Functional enrichment analysis for genes in cluster G1.
https://doi.org/10.1371/journal.pone.0311061.s005
(CSV)
S2 Table. Functional enrichment analysis for genes in cluster G2.
https://doi.org/10.1371/journal.pone.0311061.s006
(CSV)
S3 Table. Functional enrichment analysis for genes in cluster G3.
https://doi.org/10.1371/journal.pone.0311061.s007
(CSV)
S4 Table. Functional enrichment analysis for genes in cluster C1.
https://doi.org/10.1371/journal.pone.0311061.s008
(CSV)
S5 Table. Functional enrichment analysis for genes in cluster C2.
https://doi.org/10.1371/journal.pone.0311061.s009
(CSV)
S6 Table. Functional enrichment analysis for genes in cluster C3.
https://doi.org/10.1371/journal.pone.0311061.s010
(CSV)
S7 Table. Functional enrichment analysis for genes in cluster C4.
https://doi.org/10.1371/journal.pone.0311061.s011
(CSV)
S8 Table. Functional enrichment analysis for genes in cluster C5.
https://doi.org/10.1371/journal.pone.0311061.s012
(CSV)
S9 Table. Functional enrichment analysis for genes in cluster C6.
https://doi.org/10.1371/journal.pone.0311061.s013
(CSV)
S10 Table. Functional enrichment analysis for genes in cluster C7.
https://doi.org/10.1371/journal.pone.0311061.s014
(CSV)
S11 Table. Functional enrichment analysis for genes in cluster C8.
https://doi.org/10.1371/journal.pone.0311061.s015
(CSV)
S12 Table. Functional enrichment analysis for genes found in both clusters G1 and C2.
https://doi.org/10.1371/journal.pone.0311061.s016
(CSV)
S13 Table. Functional enrichment analysis for genes found in both clusters G2 and C1.
https://doi.org/10.1371/journal.pone.0311061.s017
(CSV)
S14 Table. Functional enrichment analysis for all genes showing significant non-additive interaction between the genotype and spatiotemporal factors.
https://doi.org/10.1371/journal.pone.0311061.s018
(CSV)
S15 Table. Functional enrichment analysis for genes in cluster I9.
https://doi.org/10.1371/journal.pone.0311061.s019
(CSV)
S16 Table. Functional enrichment analysis for genes in cluster I1.
https://doi.org/10.1371/journal.pone.0311061.s020
(CSV)
S17 Table. Genes in extracellular clusters E1-5.
https://doi.org/10.1371/journal.pone.0311061.s021
(CSV)
References
- 1. Palkova Z, Vachova L. Spatially structured yeast communities: Understanding structure formation and regulation with omics tools. Comput Struct Biotechnol J. 2021;19:5613–21. Epub 20211009. pmid:34712401; PubMed Central PMCID: PMC8529026.
- 2. Stewart PS, Franklin MJ. Physiological heterogeneity in biofilms. Nat Rev Microbiol. 2008;6(3):199–210. pmid:18264116.
- 3. van Gestel J, Vlamakis H, Kolter R. Division of Labor in Biofilms: the Ecology of Cell Differentiation. Microbiol Spectr. 2015;3(2):MB-0002-2014. pmid:26104716.
- 4. Reynolds TB, Fink GR. Bakers’ yeast, a model for fungal biofilm formation. Science. 2001;291(5505):878–81. pmid:11157168.
- 5. Vachova L, Stovicek V, Hlavacek O, Chernyavskiy O, Stepanek L, Kubinova L, et al. Flo11p, drug efflux pumps, and the extracellular matrix cooperate to form biofilm yeast colonies. J Cell Biol. 2011;194(5):679–87. pmid:21875945; PubMed Central PMCID: PMC3171128.
- 6. Kuthan M, Devaux F, Janderova B, Slaninova I, Jacq C, Palkova Z. Domestication of wild Saccharomyces cerevisiae is accompanied by changes in gene expression and colony morphology. Mol Microbiol. 2003;47(3):745–54. pmid:12535073.
- 7. Stovicek V, Vachova L, Palkova Z. Yeast biofilm colony as an orchestrated multicellular organism. Commun Integr Biol. 2012;5(2):203–5. pmid:22808334; PubMed Central PMCID: PMC3376065.
- 8. Marsikova J, Wilkinson D, Hlavacek O, Gilfillan GD, Mizeranschi A, Hughes T, et al. Metabolic differentiation of surface and invasive cells of yeast colony biofilms revealed by gene expression profiling. BMC Genomics. 2017;18(1):814. pmid:29061122; PubMed Central PMCID: PMC5654107.
- 9. Chow J, Starr I, Jamalzadeh S, Muniz O, Kumar A, Gokcumen O, et al. Filamentation Regulatory Pathways Control Adhesion-Dependent Surface Responses in Yeast. Genetics. 2019;212(3):667–90. Epub 20190503. pmid:31053593; PubMed Central PMCID: PMC6614897.
- 10. Cromie GA, Tan Z, Hays M, Sirr A, Jeffery EW, Dudley AM. Transcriptional Profiling of Biofilm Regulators Identified by an Overexpression Screen in Saccharomyces cerevisiae. G3 (Bethesda). 2017;7(8):2845–54. pmid:28673928; PubMed Central PMCID: PMC5555487.
- 11. Cook JG, Bardwell L, Kron SJ, Thorner J. Two novel targets of the MAP kinase Kss1 are negative regulators of invasive growth in the yeast Saccharomyces cerevisiae. Genes Dev. 1996;10(22):2831–48. pmid:8918885.
- 12. Conlan RS, Tzamarias D. Sfl1 functions via the co-repressor Ssn6-Tup1 and the cAMP-dependent protein kinase Tpk2. J Mol Biol. 2001;309(5):1007–15. pmid:11399075.
- 13. Granek JA, Magwene PM. Environmental and genetic determinants of colony morphology in yeast. PLoS Genet. 2010;6(1):e1000823. pmid:20107600; PubMed Central PMCID: PMC2809765.
- 14. Vandermeulen MD, Cullen PJ. Gene by Environment Interactions reveal new regulatory aspects of signaling network plasticity. PLoS Genet. 2022;18(1):e1009988. Epub 20220104. pmid:34982769; PubMed Central PMCID: PMC8759647.
- 15.
Rose M, Winston F. and Hieter P. Methods in Yeast Genetics—A Laboratory Course Manual. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press; 1990.
- 16. Ruusuvuori P, Lin J, Scott AC, Tan Z, Sorsa S, Kallio A, et al. Quantitative analysis of colony morphology in yeast. Biotechniques. 2014;56(1):18–27. pmid:24447135; PubMed Central PMCID: PMC3996921.
- 17. Collart MA, Oliviero S. Preparation of yeast RNA. Curr Protoc Mol Biol. 2001;Chapter 13:Unit13 2. pmid:18265096.
- 18. Cromie GA, Tan Z, Hays M, Jeffery EW, Dudley AM. Dissecting Gene Expression Changes Accompanying a Ploidy-Based Phenotypic Switch. G3 (Bethesda). 2017;7(1):233–46. Epub 20170105. pmid:27836908; PubMed Central PMCID: PMC5217112.
- 19. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat Methods. 2012;9(4):357–9. pmid:22388286; PubMed Central PMCID: PMC3322381.
- 20. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. 2014;30(7):923–30. pmid:24227677.
- 21. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2010;26(1):139–40. pmid:19910308; PubMed Central PMCID: PMC2796818.
- 22.
R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2021.
- 23. Holm S. A Simple Sequentially Rejective Multiple Test Procedure. Scandinavian Journal of Statistics. 1979;6(2):65–70.
- 24. Kolde R. Pheatmap: pretty heatmaps. R Package Version 1.0.12 ed2019. p. Implementation of heatmaps that offers more control. over dimensions and appearance.
- 25. Liis Kolberg UR, Ivan Kuzmin, Priit Adler, Jaak Vilo, Hedi Peterson. g:Profiler—interoperable web service for functional enrichment analysis and gene identifier mapping: Nucleic Acids Research; 2023. Available from: https://biit.cs.ut.ee/gprofiler/gost.
- 26. Tan Z, Hays M, Cromie GA, Jeffery EW, Scott AC, Ahyong V, et al. Aneuploidy underlies a multicellular phenotypic switch. Proc Natl Acad Sci U S A. 2013;110(30):12367–72. Epub 2013/07/03. pmid:23812752; PubMed Central PMCID: PMC3725063.
- 27. Lo WS, Dranginis AM. The cell surface flocculin Flo11 is required for pseudohyphae formation and invasion by Saccharomyces cerevisiae. Mol Biol Cell. 1998;9(1):161–71. pmid:9436998; PubMed Central PMCID: PMC25236.
- 28. Guo B, Styles CA, Feng Q, Fink GR. A Saccharomyces gene family involved in invasive growth, cell-cell adhesion, and mating. Proc Natl Acad Sci U S A. 2000;97(22):12158–63. pmid:11027318; PubMed Central PMCID: PMC17311.
- 29. Gimeno CJ, Fink GR. Induction of pseudohyphal growth by overexpression of PHD1, a Saccharomyces cerevisiae gene related to transcriptional regulators of fungal development. Mol Cell Biol. 1994;14(3):2100–12. pmid:8114741; PubMed Central PMCID: PMC358570.
- 30. Malcher M, Schladebeck S, Mosch HU. The Yak1 protein kinase lies at the center of a regulatory cascade affecting adhesive growth and stress resistance in Saccharomyces cerevisiae. Genetics. 2011;187(3):717–30. pmid:21149646; PubMed Central PMCID: PMC3063667.
- 31. Cullen PJ, Sabbagh W Jr, Graham E, Irick MM, van Olden EK, Neal C, et al. A signaling mucin at the head of the Cdc42- and MAPK-dependent filamentous growth pathway in yeast. Genes Dev. 2004;18(14):1695–708. pmid:15256499; PubMed Central PMCID: PMC478191.
- 32. Terashima H, Fukuchi S, Nakai K, Arisawa M, Hamada K, Yabuki N, et al. Sequence-based approach for identification of cell wall proteins in Saccharomyces cerevisiae. Curr Genet. 2002;40(5):311–6. pmid:11935221.
- 33. Verna J, Lodder A, Lee K, Vagts A, Ballester R. A family of genes required for maintenance of cell wall integrity and for the stress response in Saccharomyces cerevisiae. Proc Natl Acad Sci U S A. 1997;94(25):13804–9. pmid:9391108; PubMed Central PMCID: PMC28388.
- 34. Dahmann C, Diffley JF, Nasmyth KA. S-phase-promoting cyclin-dependent kinases prevent re-replication by inhibiting the transition of replication origins to a pre-replicative state. Curr Biol. 1995;5(11):1257–69. pmid:8574583.
- 35. Zhu J, Jia ZW, Xia CY, Gao XD. The Sur7/PalI family transmembrane protein Tos7 (Yol019w) plays a role in secretion in budding yeast. Fungal Genet Biol. 2020;144:103467. Epub 20200928. pmid:33002606.
- 36. Vandermeulen MD, Cullen PJ. New Aspects of Invasive Growth Regulation Identified by Functional Profiling of MAPK Pathway Targets in Saccharomyces cerevisiae. Genetics. 2020;216(1):95–116. Epub 20200714. pmid:32665277; PubMed Central PMCID: PMC7463291.
- 37. Liu H, Styles CA, Fink GR. Saccharomyces cerevisiae S288C has a mutation in FLO8, a gene required for filamentous growth. Genetics. 1996;144(3):967–78. pmid:8913742; PubMed Central PMCID: PMC1207636.
- 38. Fujita A, Hiroko T, Hiroko F, Oka C. Enhancement of superficial pseudohyphal growth by overexpression of the SFG1 gene in yeast Saccharomyces cerevisiae. Gene. 2005;363:97–104. Epub 20051109. pmid:16289536.
- 39. Kang PJ, Angerman E, Nakashima K, Pringle JR, Park HO. Interactions among Rax1p, Rax2p, Bud8p, and Bud9p in marking cortical sites for bipolar bud-site selection in yeast. Mol Biol Cell. 2004;15(11):5145–57. Epub 20040908. pmid:15356260; PubMed Central PMCID: PMC524791.
- 40. Vachova L, Palkova Z. Physiological regulation of yeast cell death in multicellular colonies is triggered by ammonia. J Cell Biol. 2005;169(5):711–7. Epub 2005/06/09. pmid:15939758; PubMed Central PMCID: PMC2171614.
- 41. Cap M, Stepanek L, Harant K, Vachova L, Palkova Z. Cell differentiation within a yeast colony: metabolic and regulatory parallels with a tumor-affected organism. Mol Cell. 2012;46(4):436–48. Epub 2012/05/09. pmid:22560924.
- 42. Piccirillo S, White MG, Murphy JC, Law DJ, Honigberg SM. The Rim101p/PacC pathway and alkaline pH regulate pattern formation in yeast colonies. Genetics. 2010;184(3):707–16. Epub 2009/12/30. pmid:20038633; PubMed Central PMCID: PMC2845339.