Gene by Environment Interactions reveal new regulatory aspects of signaling network plasticity

Phenotypes can change during exposure to different environments through the regulation of signaling pathways that operate in integrated networks. How signaling networks produce different phenotypes in different settings is not fully understood. Here, Gene by Environment Interactions (GEIs) were used to explore the regulatory network that controls filamentous/invasive growth in the yeast Saccharomyces cerevisiae. GEI analysis revealed that the regulation of invasive growth is decentralized and varies extensively across environments. Different regulatory pathways were critical or dispensable depending on the environment, microenvironment, or time point tested, and the pathway that made the strongest contribution changed depending on the environment. Some regulators even showed conditional role reversals. Ranking pathways’ roles across environments revealed an under-appreciated pathway (OPI1) as the single strongest regulator among the major pathways tested (RAS, RIM101, and MAPK). One mechanism that may explain the high degree of regulatory plasticity observed was conditional pathway interactions, such as conditional redundancy and conditional cross-pathway regulation. Another mechanism was that different pathways conditionally and differentially regulated gene expression, such as target genes that control separate cell adhesion mechanisms (FLO11 and SFG1). An exception to decentralized regulation of invasive growth was that morphogenetic changes (cell elongation and budding pattern) were primarily regulated by one pathway (MAPK). GEI analysis also uncovered a round-cell invasion phenotype. Our work suggests that GEI analysis is a simple and powerful approach to define the regulatory basis of complex phenotypes and may be applicable to many systems.

One way that phenotypes can be regulated is by signal transduction pathways. Signaling pathways receive and process stimuli from external (or internal) environments and transmit information to trigger a response, usually by influencing the expression of target genes. Signaling pathways do not operate independently, and instead function with other pathways and protein complexes in signaling networks [50][51][52][53][54]. Such networks allow the integration of diverse stimuli into a response. Signaling networks have been extensively studied by proteininteraction tests, which reveal dozens to hundreds of interacting proteins [55,56], mathematical modeling to understand network fidelity and integration [57,58], and increasingly by systems-wide approaches [59][60][61], which show the comprehensive nature of these web-like networks. Despite numerous advances in this area, many aspects of signaling network regulation remain mysterious. From the perspective of phenotypic plasticity, an interesting and important question is how do signaling networks change output phenotypes in different settings?
Invasive growth has mainly been studied under select laboratory conditions. However, S. cerevisiae is a free-living microbe that encounters diverse environments in wild [111][112][113][114] and domesticated settings [115][116][117][118][119]. We reasoned that by exploring invasive growth in different environments, we might learn more about the regulatory basis of the response. By analyzing mutants on different environments (i.e. GEI analysis) in a broad-based survey, which is a useful approach to understand gene function [120,121], we gained new insights into the invasive growth regulatory network. Specifically, we found pathways made different contributions to the regulation of invasive growth in different environments. We were able to rank pathways by their overall contribution, which led to the identification of an underappreciated pathway as one of the main regulators of the response. We also show that depending on the environment, pathways showed conditional redundancy, role reversals, cross-pathway regulation, and conditional regulation of gene expression. We also identified a round-cell invasion phenotype. The knowledge gained from this simple approach may extend to signaling networks in general. Moreover, analyzing GEIs of signaling pathways may be a straightforward approach to employ in other systems to learn more about the regulation of complex phenotypes.

GEI analysis uncovers diversity in invasive growth regulation and identifies a new major regulator
Invasive growth was examined across environments, including standard laboratory medium [YPD, yeast extract, peptone, dextrose], and media with varying amounts and/or sources of carbon, nitrogen, and phosphate, including media thought to mimic a natural environment [SOE, synthetic oak extract, [122]]. Media containing an inducer (ethanol) and inhibitor (KCl) of the response were also tested (13 environments total, S1 Table). Wild-type cells of the filamentous (∑1278b) strain background were spotted onto the different media, and invasive growth was evaluated by the plate-washing assay (PWA, [65]), where a community of cells washed off a surface leaves a visible invasive scar (Fig 1B, YPD). Invasive growth was quantified and normalized to colony size for at least three biological replicates [M&M and [123]]. Wild-type cells showed a 10-fold range in invasive growth across environments (Fig 1B and  1C, full PWA in S1A and S1B, statistical analysis in S1 File). YPD medium induced a moderate level of invasive growth (bar with cross hatch), while several environments induced robust invasion including the environment representing a natural setting (SOE). Invasive growth did not correlate with colony size (S1C Fig). These results support the existing idea that invasive growth is not binary (ON/OFF) but occurs in a phenotypic spectrum based on the environment. These properties make the invasive growth phenotype a strong candidate for GEI analysis.
The finding that invasive growth changes across environments indicates that the response may be regulated differently in different settings. Many signaling pathways, transcription factors, and protein complexes regulate invasive growth (Fig 1D, Function). To define how the regulation of invasive growth differs in each environment, the PWA was performed with mutants that perturb key pathways and protein complexes required for invasive growth (Fig  1D, Strain). Genes were chosen that ablate pathway activity based on previous tests comparing mutants that lack key pathway components [86,88,124]. The invasive growth of mutants was first compared to wild-type (Fig 1D, red, less invasive than WT; blue, more invasive, white, no difference, raw data S1 Fig, quantitation with standard deviation in S2 Fig, statistical analysis in S2 File). Most mutants showed differences in invasive growth from wild type (Fig 1D, e.g. YPD, tenth row, ras2Δ mutant showed decreased invasion relative to wild type, p-value <0.0001). Mutants also showed differences across environments (Fig 1D, compare colors across rows, statistical analysis in S1 File). For example, the ras2Δ mutant showed variable shades of red across environments (Fig 1D, e.g. compare YPGAL to WL, p-value < 0.0001, YPGAL to SLAD, p-value 0.04, and WL to SLAD, p-value 0.02). The mutants were also different from each other (Fig 1D, compare colors down columns, statistical analysis in S3 File). The ras2Δ mutant for example was not as critical as the opi1Δ mutant for invasive growth on YPGAL medium (Fig 1D, ras2Δ is a lighter shade of red, p-value 0.0006). These results demonstrate that the regulation of invasive growth changes based on the environment.
Because the role each pathway played in regulating invasive growth changed across environments, we asked whether some pathways play a major role in more environments than others, making them a stronger overall regulator of the biological process. To define the signaling pathways with the strongest overall role in regulating invasive growth, we ranked the regulators based on their roles in all environments collectively. We first averaged the invasive growth of each mutant relative to wild type across environments and arranged them in ascending order (S3A Fig). Because average relative invasion can have high variability, we also summed the number of environments that each mutant showed a specified reduction in invasion compared to wild type for several thresholds (S3B Fig). Both methods showed a different overall ranking of the regulators compared to the order seen in YPD (compare the order in Figs 1D to S3). Unexpectedly, both methods identified the transcriptional regulator Opi1p as the strongest single regulator of invasive growth (Fig 1E, opi1Δ). Opi1p is a positive regulator of invasive growth and biofilm/mat formation [125], and functions as a transcriptional repressor of phospholipid biosynthesis [126][127][128], but has not been appreciated as a major regulator of the response. Ranking also showed three well-established regulators of invasive growth among the top pathways: RIM101 (by rim101Δ), MAPK (by tec1Δ), and RAS (by ras2Δ) [Fig 1E]. Because invasion phenotypes can be influenced by genetic background [129][130][131], our ranking may be specific to the strain background (S1278b) and environments tested here. The ability to order pathways allowed us to focus on the most relevant regulators. Thus, GEI analysis allowed ordering pathways that regulate invasive growth and identified an under-appreciated transcription factor as the major regulator of the response.

Plasticity in invasive growth regulation revealed by GEI analysis between environments, microenvironments, and over time
To further explore variations in invasive growth regulation, strains lacking the main regulators (RAS, MAPK, RIM101, and OPI1), which showed no growth defects compared to wild-type cells (S4 Fig), were re-examined across environments. As shown above, OPI1 was a major regulator in most environments (Fig 2A-2C, YPD, YPGAL, SLPD, WL, larger circle represents a stronger contribution). However, in one environment (YP), OPI1 had a minor role (S1 File, pvalue < 0.0001). The idea that a pathway can play a major role in some environments but not others was also true for RIM101 (Fig 2A-2C, compare YP to WL, p-value 0.0001), RAS (Fig  2A-2C, compare WL to YPD, p-value 0.0005), and MAPK (Fig 2A-2C, compare YPGAL to SLPD, p-value < 0.0001). These results reveal a highly plastic regulatory network, where no single regulator has centralized control, and where different pathways play the major role in different environments.
We noticed when comparing two environments where wild type showed a change in invasion, some mutants showed the same change while other mutants showed no change. For example, wild-type cells showed increased invasion on WL medium compared to SLPD medium (Figs 2A and 1C, statistics in S1 File, p-value 0.0002) as did the rim101Δ mutant (Fig  2A and 2B, p-value 0.0005); however, the opi1Δ mutant showed a similar level of invasion (Fig  2A and 2B, p-value, n.s., 0.78). These data may suggest when a pathway is involved in sensing the difference between the environments (i.e. OPI1 in this example) compared to a non-sensing role (i.e. RIM101 in this example). Future studies aimed at testing these predictions by transitioning cells from one environment to another may uncover sensory mechanisms in the regulation of invasive growth.
Cells growing within microbial communities experience different microenvironments. Microenvironments can trigger specific responses in subsets of cells throughout the community [132]. Invasive growth occurs in distinct patterns, reflecting different microenvironments, that can be quantified and represented in a plot profile [123]. Plot profiles generate a representative cross section where less invasion (light pixel intensity) is represented by high values, and more invasion (dark pixel intensity) is represented by low values (Fig 2D and 2E, e.g. SGAL, wild-type, black line). Wild-type cells showed different invasion patterns in different environments (Fig 2D and 2E, black line, top and bottom panels). We examined how the major pathways might affect the regulation of invasive growth in different microenvironments. Cells lacking the main regulatory pathways also showed different patterns. For example, in one environment (SGAL), the tec1Δ mutant showed the highest invasion in the center of the scar, and the ras2Δ mutant showed the highest invasion in a surrounding ring (Fig 2D and 2E, SGAL, compare green and grey lines). This indicates that in SGAL medium, RAS is critical for invasion in the center region, and MAPK is critical in the surrounding ring region. In a different environment (SOE), the role of the regulatory pathways changed. On SOE, the ras2Δ mutant showed the most invasion in the center region, and the tec1Δ mutant showed a diffuse pattern (Fig 2D and 2E, SOE). This means that in some environments, the MAPK pathway is more critical than RAS for invasion in the center region (see also YPGAL, S5 Fig). These results support the idea that different pathways take on the major role in regulating invasive growth based on the environment.
Many pathways and effector genes in the invasive growth network also play a role in regulating complex-colony morphology [ruffling of cells above the surface [133]], which is an aspect of biofilm/mat formation [134]. Biofilm/mat formation also occurs in distinct patterns resulting from different microenvironments [135]. Wild-type colonies had a ruffled pattern on some environments (Fig 1A, YPGAL) but not others (Fig 1A, SOE). The different patterns did  S1 Fig. B) Levels of relative invasion to wild type, with wild type values set to 1. Asterisk, P-value � 0.05, compared to wild type. C) Model of pathway contributions to invasive growth. Circle size is proportional to pathway contribution. D) PWA, see panel A for details. E) Plot profile of invasive growth across invasive scars. Strains and not correspond to differences in invasive growth (Fig 1B, compare YPGAL to SOE, S1 File, pvalue 0.0015). This result indicates that different regulatory mechanisms occur above and below the surface. Many pathways played a similar role in generating phenotypes above and below the surface (S1 Fig); however, at least one pathway played different roles in these microenvironments (S6 Fig, RTG). This result supports the idea that biofilm/mat formation and invasive growth are regulated by both overlapping and non-overlapping sets of pathways/ genes [73]. Therefore, microenvironments reveal plasticity in the roles pathways play in regulating filamentous growth in different regions within a community of cells.
We became interested in understanding whether the pathways also differentially regulate invasive growth at different time points. To address this possibility, invasive growth was examined over time. In one environment (YPGAL), invasive growth increased over a three-week incubation (Fig 2F, full PWA in S7A). All four pathways played a major role in regulating invasive growth at an early time point tested (2d) but played different roles at different times ( Fig  2F and 2G, YPGAL, 3d-21d). For example, MAPK was more critical at an early time point (3d) compared to a later time point (21d) for invasive growth (Fig 2F and 2G). Interestingly, the pathways made different contributions from one another at earlier time points (2d-7d) but made the same overall contribution at the last time point tested (21d) [Fig 2G, YPGAL]. However, the patterns were different (Fig 2F, YPGAL). Therefore, major regulatory pathways can make varying contributions during the progression of invasive growth.
GEI analysis was also used to explore the temporal development of invasive growth across environments. Like for YPGAL medium, on SD medium invasive growth increased over time (Figs 2F and S7B). However, pathways played different roles over time in the two conditions, based on the levels and patterns of invasive growth (Fig 2F and 2G, SD, and S7C). For example, when comparing 7d to 21d, RIM101 contributed similarly to invasive growth at both time points in one environment (YPGAL) but differently at both time points in another environment (SD) [Fig 2G]. Moreover, all four pathways did not equally contribute by 21d on SD like they did on YPGAL medium (Fig 2G). Invasive growth after 21d also revealed GEIs for invasive growth patterns not seen at earlier time points between YPGAL and SD media (Fig 2F, 21 d, e.g. RIM101 was required for center-ring invasion on SD but not YPGAL). Thus, the temporal roles pathways play in the development of the response may change in different environments.

Additive and redundant roles of invasive growth regulators
When regulating a phenotype, signaling pathways can make separate (i.e. additive) contributions from one another, or they can make overlapping (i.e. redundant) contributions. To test if the major pathways played additive or redundant roles during invasive growth, double mutant combinations were generated among the major regulators. In most environments, the double mutants were less invasive than the single mutants. For example, on YPD, the ras2Δ tec1Δ double mutant was less invasive than the ras2Δ and tec1Δ single mutants [Fig 3A and 3B, full data set in S8)]. This indicates that the pathways play additive roles.
In some environments, double mutants showed the same invasion as single mutants. For example, the ras2Δ tec1Δ double mutant had the same level of invasive growth as the ras2Δ and tec1Δ single mutants on YP (Fig 3A and 3B). Also, the ras2Δ opi1Δ double mutant was less invasive than either single mutant in one environment (YPD), equal to one of the single media are as indicated; examples are shown in panel D. X-axis, distance (in pixels); Y-axis, pixel intensity. F) Invasive growth over time. PWA was performed at the times indicated. See S7 Fig for raw  mutants in multiple environments (opi1Δ on YPGAL, SOE, WL), and equal to both single mutants in yet another environment (YP) [Figs 3A, 3B and S8]. This may suggest that a pathway in the network can make an additive contribution in one environment (Fig 3C, RAS on YPD), a redundant or similar contribution to one pathway in another environment (Fig 3C, RAS on YPGAL), and a redundant or similar contribution to more than one pathway in a third environment (Fig 3C, RAS on YP). In environments where double mutants invade, it can be inferred that at least one additional pathway is required. These observations are consistent with the idea that pathways that regulate invasive growth can play additive or redundant roles depending on the environment, which we refer to as conditional redundancy.

Conditional role reversals between pathways can achieve similar invasive states
GEI analysis showed that several pathways promoted invasive growth in one environment and inhibited it in another (Fig 1C, RTG, PHO85, and other pathways, red or blue in different environments). These included Retrograde (RTG), which regulates the response to mitochondrial distress [136,137] and invasive growth [89,138,139]; and the alternate cyclin-dependent kinase (CDK) Pho85p, which controls cell-cycle progression in certain environments [140][141][142][143] and invasive growth [88]. These 'conditional role reversals' were explored in more detail. RTG was required for invasive growth in one environment (Fig 3D and 3E, YPD, rtg3Δ) and inhibited invasive growth in another (Fig 3D and 3E, SLPD). In some environments, RTG played no role (Fig 3D and 3E, YP). Pho85p was required for invasive growth in SLPD and inhibited it in SD (Fig 3D and 3E, pho85Δ). Intriguingly, when comparing SLPD to SD media, the environment where RTG played a negative role, Pho85p played a positive role and vice versa (Fig 3F). Despite these and other regulatory differences (S9 Fig), the amount and pattern of invasive growth for wild type was similar in both environments. Thus, wild-type cells can exhibit a similar phenotype between environments despite opposing roles for network regulators. RTG and Pho85 might be functionally connected such that in the absence of one pathway, the other takes on a different role. In any case, this finding supports the idea that the same phenotype can arise through the action of pathways that have opposing roles in different settings.

Signaling pathways conditionally regulate MAPK pathway activity across environments
Signaling pathways function in basal and activated states. In activated states, pathways can show different kinetics (amplitude, duration), which can result in different responses [ERK is a classic example [144]]. To define how the activity of the MAPK pathway might change across environments, wild-type cells containing a translational reporter whose activity reflects pathway activity [pFRE-lacZ, [89,145]] was examined. Based on reporter activity, the MAPK pathway showed a >10-fold range in activity across environments (Fig 4A). MAPK pathway activity correlated with invasive growth (S7D Fig). In line with previous findings [146], this result demonstrates that the MAPK pathway does not operate in an "ON/OFF" state but in a graded manor depending on the environment.
The levels of the reporter in cells lacking a positive regulator of the MAPK pathway defined a lower boundary for MAPK pathway activity (Fig 4B, tec1Δ, green). Reporter levels in cells lacking a negative regulator of the MAPK pathway, Dig1p [147][148][149][150], defined an upper boundary (Fig 4B, dig1Δ, purple). MAPK pathway activity was closer to the lower boundary in the environments tested (Fig 4C, MAPK pathway activity). Like MAPK pathway activity, MAPKdependent invasive growth in wild-type cells was closer to the lower boundary (represented by the hypoinvasive tec1Δ mutant) compared to the upper boundary (represented by the hyperinvasive dig1Δ mutant [S1 and S2B Figs]) in most environments tested, based on examination of invasive growth across conditions (Fig 4C, Invasive Growth) and over time (Figs 4D and S7E). One exception was on YP media, where invasive growth, but not MAPK pathway activity, was closer to the upper boundary (Fig 4C). This could be due to redundancy with another pathway that regulates invasive growth in this environment (i.e. RAS). Although an environment that triggers maximal invasive growth and MAPK pathway activity may remain to be identified, these results suggest that the MAPK pathway typically operates at less than half its maximum levels of activity, and promotes invasive growth to less than half its maximal levels, even in MAPK pathway activity is curbed and is conditionally regulated by other major invasive growth regulators. A) ß-galactosidase (lacZ) assay. Wild-type cells were quantified for MAPK pathway activity by the pFRE-lacZ reporter in indicated media. Values were normalized to YPD, which was set to 1. B) ß-galactosidase (lacZ) assay. Wild-type cells and the tec1Δ and dig1Δ mutants were quantified for relative MAPK pathway activity. Wild-type values were set to 1. Top panel, tec1Δ, green. Bottom panel, dig1Δ, purple. Asterisk, p-value � 0.05, compared to wild type. C) Graph of MAPK pathway activity or Invasive Growth for wild-type cells compared to the tec1Δ (set to zero) and environments where the pathway is activated to stimulate invasive growth. Thus, the MAPK pathway functions at modest levels to induce invasive growth.
Signaling pathways can regulate each other's activity [151][152][153][154][155][156]. In the filamentous/invasive growth network, RAS [78,89], RIM101 and OPI1 [89], regulate the activity of the MAPK pathway. Therefore, we became interested in understanding whether these pathways regulate MAPK pathway activity differently across environments. To test if the role in cross-pathway regulation of RAS, RIM101, or OPI1 changes, MAPK pathway activity was examined in pathway mutants across environments. Each pathway conditionally regulated MAPK pathway activity (Fig 4E and 4F, arrow size represents strength of contribution). For example, in one environment, all three pathways regulated the MAPK pathway (Fig 4E and 4F, SGAL). In other environments, the pathways did not: on YPD medium RAS was not required (Fig 4E  and 4F); on YP medium, OPI1 was not required (Fig 4E and 4F); and on SD medium, RIM101 was not required (Fig 4E and 4F). The fact that RAS, RIM101, and OPI1 contribute to MAPK pathway activity in some environments but not others demonstrates that environment controls cross-pathway regulation in the signaling network. This offers one mechanism for how MAPK pathway activity (and possibly invasive growth) exhibits plasticity. Furthermore, conditional cross-pathway regulation may account for how conditional redundancy occurs between the RAS and MAPK pathways (Fig 3C).
MAPK pathway activity might be conditionally regulated by altering the expression of the genes encoding pathway components. The sensor of the pathway, MSB2, the kinase of the pathway, KSS1, and the transcription factor, TEC1, show different expression patterns in different environments [83,86]. Moreover, global gene expression analysis and protein-DNA interaction tests showed RIM101 regulates MSB2 [89] and TEC1 expression [157]; OPI1 regulates MSB2 [89] and KSS1 expression [157]; and RAS regulates MSB2 [84,85,89] and TEC1 expression [84] (S2 Table). Other pathways might regulate each other's activities as well. The expression of the RAS2 gene is regulated by MAPK [158], RIM101 [159], and OPI1 [157]. The expression of the gene encoding the RAS pathway transcription factor, FLO8, is regulated by MAPK [85,160] and OPI1 [161]. The expression of the RIM101 gene is regulated by RAS [85,157] and the expression of RIM8 is regulated by MAPK [89,160]. The expression of OPI1 is regulated by RIM101 [159]. Thus, pathways within the network may regulate each other's activities by regulating genes encoding pathway components. The important conclusion from this section is that GEI analysis can reveal new insights into the activity and regulation of signaling pathways that function in an interconnected network.

Pathways differentially regulate separate cell adhesion target genes
Upstream signaling events induce three well-established effector phenotypes during filamentous growth: an increase in cell-to-cell adhesion, a switch to distal budding, and an increase in cell length [64,65,92]. The three effector phenotypes have some independence from each other in regulating invasive growth [162,163]; however, whether adhesion, distal budding, and cell elongation change their contribution to invasion across environments is unknown.
To explore how effector processes contribute to invasive growth in different environments, mutants that compromise each aspect [flo11Δ (reduced adhesion), bud8Δ (reduced distal dig1Δ (set to 1.0) mutants. The average MAPK pathway activity or Invasive growth of wild-type cells across environments is marked (Avg arrow). D) Same as panel C, except for invasive growth at indicated time points (2d, 3d, 7d, 21d). E) ß-galactosidase (lacZ) assay. Indicated strains were quantified for relative MAPK pathway activity, with wild type values set to 1. Asterisk, p-value � 0.05, compared to wild type. pvalue, 0.086 for the opi1Δ mutant in YPGAL and 0.07 for the rim101Δ mutant in SLPD. F) Pathway contributions to MAPK pathway activity in different environments. Arrow size is proportional to pathway contribution.
https://doi.org/10.1371/journal.pgen.1009988.g004 budding), and pea2Δ (reduced elongation)] were tested by the PWA. We found each aspect of invasive growth contributed differently from one another, in agreement with previous studies [162,163], and their contribution changed based on the environment (Fig 5A, full data set in S1-S2, statistical analysis in S1-S3 File). FLO11-dependent adhesion played a major role in invasive growth in most environments (Fig 5A-5C), but not all (YPGAL). Similarly, distal budding and cell elongation played major roles in several environments (Fig 5A-5C, e.g. YPGAL and SOE) and minor roles in others (e.g. YP and SGAL). Each aspect behaved differently than the other aspects in at least one environment (Fig 5D, PEA2 on YPD, FLO11 on YPGAL, BUD8 on +EtOH). The flo11Δ, bud8Δ, and pea2Δ mutants also showed different patterns of invasion. On SOE medium, for example, the flo11Δ mutant showed pinpoint invasion, while bud8Δ and pea2Δ mutants showed uniform invasion (Fig 5B). Therefore, one explanation for the diversity in invasive growth induced by the signaling network comes from differences in the roles of effector processes.
Because adhesion, distal budding, and elongation affect invasive growth differently, we hypothesized the differential regulation of these aspects by signaling pathways could account for how the regulatory network controls the variation in invasive growth seen in different environments. To examine the roles of pathways in regulating cell adhesion independent of the other aspects of filamentous growth, the size of groups or clumps of adhesive cells in pathway mutants was measured during growth in liquid cultures as described [123]. Signaling pathway mutants showed reduced adhesion compared to wild type and different levels from each other (Fig 5E and 5F and S10). For example, RIM101 was the main regulator of adhesion in YPGAL and SOE medium compared to the other pathways (Fig 5F). Furthermore, the pathways showed different levels of adhesion across environments. For example, OPI1 played a stronger role in SD than YPD media (Fig 5F). Thus, signaling pathways show differences in cell adhesion in different contexts, which suggests the pathways differentially regulate cell adhesion based on the environment. This conclusion was supported by examination of complex colony morphology (S11A Fig), which is also regulated by the same pathways and Flo11p [133,135,164]; within-colony adhesion [by a method we developed [123], S11B Fig]; and adhesion of cells to plastic (S12 Fig), another Flo11p-dependent adhesion phenotype [134,165].
MAPK, RAS, RIM101, OPI1, and other pathways regulate FLO11 expression [79,82,84,85,92,125], which has one of the largest promoters in the genome and is a site at which multiple transcription factors converge. Not surprisingly, the tec1Δ, rim101Δ, ras2Δ, and opi1Δ mutants showed reduced FLO11 expression by quantitative reverse transcription PCR (RT-qPCR) analysis (Fig 5G, left, YPD). In line with our hypothesis, the pathways regulated FLO11 expression to different levels from each other. In YPD, RIM101 and OPI1 played the major roles (>10-fold), followed by MAPK (3.3-fold), and RAS (1.3-fold). One pathway (RAS) showed a GEI in regulating FLO11 expression across environments (Fig 5G and 5H, compare 2-fold change in YPD to YPGAL).
Unexpectedly, some pathways played a larger role than Flo11p in several environments. The tec1Δ and opi1Δ mutants showed less invasion (e.g. YPGAL, S1 and S2 Figs, S3 File) and the tec1Δ, ras2Δ, rim101Δ, and opi1Δ mutants were less adherent (Fig 5E and 5F) than the flo11Δ mutant in several environments. These data reveal a Flo11p-independent adhesion mechanism that is regulated by pathways in the network. We previously showed that the MAPK pathway regulates the expression of SFG1 [83,86,160], which encodes a transcription factor involved in the regulation of filamentous growth [166]. Sfg1p, along with other transcription factors [167][168][169], regulates daughter-cell separation [170], has a minor role in regulating the expression of FLO11 [123], and mainly contributes to invasive growth by an adhesion mechanism that is separate from Flo11p [123]. Along with MAPK, the RAS, RIM101, and OPI1 pathways regulated SFG1 gene expression (Fig 5G, right, YPD). Thus, the main pathways that regulate filamentous  Fig 1B for details. B) PWA; Inverted images of invasive scars after wash, bar, 0.5 cm. Colonies prior to wash in S1 Fig. C) Levels of relative invasion, with wild type values set to 1. Asterisk, p-value � 0.05, compared to wild type. D) Model of effector contributions to invasive growth. Circle size is proportional to effector contribution. E) Cell adhesion in liquid cultures. Cells were imaged by microscopy at 5X magnification, bar, 200 μm. All environments in S10 Fig. F) Quantification of relative adhesion by average area of cell clusters, normalized to wild type, set to 1. Asterisk, p-value � 0.05, compared to wild type. G) RT-qPCR analysis of mRNA levels (left, FLO11; right, SFG1) between indicated strains and environments. Wild-type values were normalized to ACT1 expression and set to 1. Black asterisk, p-value < 0.05, compared to wild type. Grey asterisk, p-value < 0.05, for FLO11 expression in the ras2Δ mutant between YPD and YPGAL. Blue asterisk, p-value < 0.05, for SFG1 expression in the opi1Δ mutant between YPD and YPGAL. H) Model of pathway contributions to SFG1 and FLO11 gene expression. Arrow size is proportional to the pathway contribution. growth regulate this cell adhesion mechanism. Support for this conclusion comes from largescale analysis showing that RIM101 regulates SFG1 expression through the transcription factor Rim101p by gene expression profiling [171] and chromatin immunoprecipitation (CHIP) analysis [172]]. OPI1 also regulates SFG1 by the transcription factors Ino2p by microarray analysis [158] and Ino4p by CHIP [173]. RAS regulates SFG1 expression [83,86], perhaps through the transcription factors Flo8p, Sok2p, and Phd1p, which bind by CHIP [84].
SFG1 (like FLO11) showed changes in expression by signaling pathways across environments. In particular, when considering the mutant analysis, OPI1 showed a > 1.7-fold change between YPD and YPGAL media in regulating SFG1 gene expression (Fig 5G and 5H). Comparing target gene expression also showed that RIM101 and OPI1 were more critical for FLO11 expression compared to SFG1 (Fig 5G and 5H). By comparison, RAS was more critical for SFG1 expression than FLO11 (Fig 5G and 5H), and the MAPK pathway showed equal regulation of both (Fig 5G and 5H). These results collectively demonstrate that the main pathways that control invasive growth control the expression of target genes that control separate cell adhesion mechanisms. Other targets that regulate invasive growth are most likely changing conditionally, because some MAPK pathway targets are regulated by the MAPK pathway in one environment but not another (S13 Fig). Such regulation may account for the phenotypic plasticity of invasive growth.

The MAPK pathway is the main regulator of cell elongation and distal budding during invasive growth
Like for adhesion, the roles pathways play in regulating cell elongation and distal budding may also differ between pathways. The MAPK pathway is known to regulate cell elongation by extending the G1 phase of the cell cycle [174] by regulating expression of CLN1 [160,175], which results in prolonged growth at the tip of the cell. The role of the other pathways in regulating cell elongation have not been fully explored. Measuring the length of filamentous cells showed that RIM101 and OPI1 played no role in regulating cell elongation, RAS played a minor role, and MAPK played the major role (Fig 6A, L/W; see additional environments in S14). A similar result was obtained for distal-pole budding: MAPK was the major pathway to regulate distal budding, while the other pathways played minor roles (Fig 6A, CFW, DB%). These results indicate that the MAPK pathway, and to a lesser degree RAS, regulate differentiation to the filamentous cell type.
The fact that strains that cannot produce elongated cells exhibit invasive growth on most environments (S1 and S2 Figs, tec1Δ and pea2Δ) indicates that cell elongation is dispensable for invasive growth. Previous reports have suggested that cells with a low level of filamentation can exhibit invasive growth [176]. To further investigate this phenotype, the invasive scar of the tec1Δ mutant was inspected by microscopy. Like for wild-type cells, the tec1Δ mutant invaded below the agar surface (Fig 6B, close up, white arrows) despite the fact that cells were round in appearance (Fig 6B, close up, compare yellow arrows). Cross sections of the invasive scar showed sub-surface invasion (Fig 6B, cross sections, full imaging in S15) and agar squashes of embedded invasive cells revealed a round-cell morphology (S16 Fig). GEI analysis can therefore uncover new phenotypes, like the round-cell invasion phenotype described here.

Discussion
Here we employed GEI analysis in a model genetic organism to explore the regulation of a complex phenotype. We hypothesized new insights would be gained by exploring several environments because invasive growth is typically studied under select laboratory conditions. Our approach uncovered new insights into the network that regulates invasive growth. It allowed us to rank the pathways in the network and assign an underappreciated regulator as a new major regulator; it uncovered conditional role reversals for some regulators; it revealed new aspects of regulation occurring between pathways in the network, such as conditional pathway redundancy and cross-pathway regulation; it revealed the conditional regulation of gene expression for downstream targets of the signaling network; and it identified a new phenotype (round-cell invasion). Given that many of the signaling pathways that regulate invasive growth are evolutionarily conserved, these finding may be broadly applicable to signaling networks in general.
A specific benefit to GEI analysis is the ability to rank the contributions of regulatory pathways to a biological response. For the filamentous/invasive growth network, GEI analysis allowed identification of the major regulators distinguishing them from pathways that contribute only under select environments. Ranking led to the unexpected discovery of an underappreciated transcriptional repressor (OPI1) as one of the major regulators of the response. Opi1p is a transcriptional repressor of phospholipid biosynthesis and inositol production [126,127,177,178]. As such, OPI1 controls the switch between lipid storage, which occurs during fasting, and membrane production, which occurs during growth [179,180]. Opi1p is inactivated by binding directly to phosphatidic acid [128,181,182], which is a conserved signaling molecule in plants and mammals [183][184][185][186] and whose misregulation is associated with cancer and other diseases [187][188][189][190][191][192]. Thus, analogous phospholipid regulatory circuits may regulate eukaryotic regulatory networks to control morphogenetic pathways (e.g. MAPK) in healthy and disease states.
GEI analysis also allowed interrogation of the flexible and invariant components of the network and showed that the major pathways (MAPK, RAS, RIM101, and OPI1) had different roles in different environments, microenvironments, and over time. Most pathways were at least partially dispensable under one or more environments revealing unexpected network plasticity. Several pathways showed conditional redundancy with other pathways. Conditional redundancy implies pathway interactions within the network are conditional and may apply to all types of proteins in networks. Perhaps this may suggest redundancy in protein function is not conserved for robustness or genetic buffering, but instead is conserved because nonredundant functions exist in some environments. In line with the finding of conditional redundancy, cross-pathway regulation was conditional: RAS, RIM101, and OPI1 regulated MAPK pathway activity only under select environments. Collectively, these results indicate that none of these pathways function as the central regulator and that the lead role is environmentally determined. Decentralizing control of invasive growth could be advantageous because it allows multiple pathways that sense different stimuli to take the leading role in different settings.
Surprisingly, conditional role reversals occurred for some pathways that control invasive growth. In these instances, GEI analysis reduces possible mischaracterization of a pathway's role due to the exploration on multiple environments. For example, if RTG was only explored on one condition (e.g. SLPD), it might be labeled as a negative regulator, even though in most environments, that pathway plays a positive role. There is precedent for transcription factors switching from positive to negative roles, which can occur in different nutrient states [193]. However, here we show that two pathways (RTG and PHO85) play opposing roles under Yellow arrows, show elongated morphology for wild type compared to round-cell morphology in the tec1Δ mutant. Cross sections, top row, 5X magnification by microscopy of invasive scar cross section, bar, 500 μm. Black arrows, agar surface. Red arrows, invading cells. Bottom row, 10X magnification by microscopy of invasive scar cross section, bar, 100 μm. Red arrows, same invading cells as 5X magnification.
https://doi.org/10.1371/journal.pgen.1009988.g006 different conditions to create a similar phenotype. How this type of regulation occurs is not clear. It will be interesting to see if conditional role reversals occur for one or all phenotypes a pathway regulates.
The fact that MAPK pathway activity functions at modest levels to induce invasive growth is consistent with previous results from our laboratory. The MAPK pathway induces expression of genes encoding multiple negative regulators of invasive growth and a negative regulator of MAPK pathway activity (NFG1), which has the effect of modulating the response [123]. Curbing invasive growth is advantageous in the balance between invasion and colony expansion [100]. Curbing MAPK pathway activity may be advantageous in that pathway hyperactivation leads to defects in budding and cell morphology [88,146,194]. Perhaps other pathways operate at low overall levels as a safeguard to curb pathway activity. It is clear at least for the mammalian RAS-MEK-ERK pathway that hyperactivation can lead to cancer and other diseases [195].
Many pathways independently converge on the FLO11 promoter. We show here that SFG1 is regulated by the major filamentation regulatory pathways as well. At least two pathways conditionally regulate the SFG1 and FLO11 promoters, which suggests that each of the major pathways independently regulates each adhesion target. Combinatorial control by multiple pathways regulating multiple targets that control adhesion by independent mechanisms can explain much of the diversity in invasive growth seen across environments. It will be interesting to determine how transcription factors bind to the FLO11 and SFG1 promoters, and we predict, based on GEI analysis, that they may bind to these promoters independently of each other and differently at the different promoters. This could occur by the transcription factors having different binding affinities or by differences in chromatin structure/organization between the promoters. The importance of cell adhesion in the regulation of filamentous/invasive growth can be seen from the existence of multiple adhesion mechanisms and the growing complexity of the adhesion code. The regulatory network may also regulate the expression of other adhesion molecules [FLO1, FLO10, FLO9, FLO5 [163,[196][197][198]] and cell-wall degrading enzymes that control cell separation [SCW11, DSE1/2/4, CTS1, ENG1 [167][168][169]199]] in an environment-dependent manner.
An exception to decentralized control of the regulatory network is that the MAPK pathway is the central regulator of polarity and cell shape. Because differentiation requires alterations to the cell cycle to trigger cell elongation [174], it may be beneficial to employ a single pathway as the main regulator. This finding led to the identification of a round-cell invasion phenotype. Round-cell invasion could have important ramifications to understanding virulence if conserved in pathogenic species. Although filamentous growth is essential for virulence in Candida albicans and Candida glabrata [200][201][202][203][204][205], how cells penetrate the host is not completely understood. Moreover, given that C. albicans contains even more cell adhesion molecules than S. cerevisiae, it may be instructive to use GEI analysis to uncover new facets of this dimorphic fungal response.
Filamentous growth was regulated at different levels in microenvironments across the invasive scar and above and below the agar surface. This may reflect differences in pH or glucose concentrations in the different environments, which exist in gradients from the center of a community of cells to the outside rim [135]. This conclusion could have important ramifications, because microenvironments play a role during infections by C. albicans, where intruders move from the site of infection to new tissues [206][207][208][209][210]. Microenvironments also play a major role in diseases like cancer [57, [211][212][213], including cancer cell invasion into new environments [214]. Intriguingly, it has recently been speculated that tumor cells encountering a new microenvironment might alter cancer progression independent of additional mutations [215]. Therefore, GEI analysis may be a useful tool to interrogate signaling networks that control complex phenotypes involved in disease and infection.
In summary, GEI analysis of a complex phenotype has revealed unexpected features of plasticity in an evolutionarily conserved regulatory network. Insights uncovered here may apply to orthologous pathways in fungal pathogens and to eukaryotic signaling pathways in general. GEI analysis can reveal new insights (and new phenotypes) by capturing the different environmental roles of regulatory pathways and preventing the characterization of only environmentspecific functions. The knowledge gained from applying this simple approach demonstrates the advantages of looking at biological systems across diverse environments and suggests this approach will be useful in interrogating other systems where signaling networks govern complex phenotypes.

Yeast strains and plasmids
Yeast strains are listed in S3 Table. All strains are isogenic with HYL333 of the ∑1278b background [provided by G. Fink [76]]. Homologous recombination was used to make gene deletions using auxotrophic/antibiotic resistance markers amplified by polymerase chain reaction (PCR). Templates were introduced into yeast by lithium acetate transformation as described [216]. PCR southern analysis and phenotype, when possible, were used to verify strains. pFRE-lacZ plasmid is used to measure the transcriptional level of the MAPK pathway as described in [217].

Microscopy
A Zeiss Axioplan 2 microscope was used for differential interference contrast (DIC) imaging. The Axiocam MRm camera was used to acquire digital images. For analysis, Axiovision 4.4 software was used. For microscopy imaging, the following strains were used: wild type (PC538) and the opi1Δ (PC2847), tec1Δ (PC569), rim101Δ (PC2953), ras2Δ (PC562), and dig1Δ (PC3039) mutants. To measure cell length-to-width ratios, images of cells were taken at 100X magnification and the images were imported into the program GIMP (https://www. gimp.org/downloads/). The measure tool was used to measure the length and width of each cell, and the larger value (length) was divided by the smaller value (width). At least 25 cells were measured per strain and the average ratio was reported with the standard deviation indicated. Budding pattern was visualized and quantified as previously described [101,221]. Fluorescent brightener #28 or Calcofluor white (CFW) was used to stain bud scars. Cells were observed at 100X magnification, and at least 100 buds per strain were counted. Large clumps of cells were excluded to avoid ambiguity. Buds were considered as emerging from either distal, proximal, or equatorial sites and tallied. The number of distal buds was divided by the number of total buds, and this value was reported as a percentage.

Plate-washing assay
The PWA was used to visualize differences in invasive growth across strains [65,222]. The following strains were used when indicated: wild type (PC538) and the opi1Δ (PC2847), tec1Δ , and pea2Δ (PC551), tec1Δ rim101Δ (PC7680), tec1Δ opi1Δ (PC7679), opi1Δ rim101Δ (PC7681), ras2Δ tec1Δ (PC7678), ras2Δ opi1Δ (PC7689), and ras2Δ rim101Δ (PC7690) mutants. The rtg3Δ (PC7677), sin3Δ (PC5864), and pbs2Δ (PC368) mutants were used on SLAD and +EtOH media and compared to wild type (PC313). Cells were spotted on indicated medium and placed at 30˚to grow for 7 d unless otherwise indicated. To ensure uniform growth, spots were placed an equal distance to each other and from the edge of the plate. To assess invasive growth, plates were placed under a stream of water, and cells were rubbed gently by hand to remove non-invasive cells. Images of colonies and the invasive scars were taken by ChemiDoc XRS+ molecular imager (Bio-Rad laboratories) under blot/chemicoloric setting with no filter. For close up images of colonys when analyzing complex-colony morphology, a Nikon D3000 digital camera was used.
Invasive growth quantitation has been described [123]. Images were imported into ImageJ (National Institutes of Health, Bethesda, MD, USA; https://imagej.nih.gov/ij/) to remove the background signal. Images were imported into the program GIMP, where images were inverted and adjusted for brightness and contrast. Each image was treated with the same parameters for adjustments. Then, as an adjustment to our previous method, images were imported into the Image Lab 6.0.1 program (Bio-Rad laboratories, https://www.bio-rad.com/ en-us/product/image-lab-software?ID=KRE6P5E8Z) for quantification using the round volume tool. Colony borders were outlined to determine area. The circle demarking the colony area was then placed over the image of the washed invasive scar, and the intensity of the invasive scar was determined by the volume tool. Next, the intensity values of the invasive scar was divided by colony area to control for growth differences between strains [intensity / area = invasion]. Invasion values were averaged across three biological replicates. To determine relative invasion, invasion was normalized to wild type, set to 1. Error bars represent standard deviation between trials. Significance was determined using the program GraphPad Prism (commercially available: https://www.graphpad.com/) by one-way ANOVA analysis unless otherwise indicated. Dunnett's multiple comparisons test was used for comparisons of mutants to wild type. Tukey's multiple comparisons test was used to compare mutants to mutants or a mutant across environments. Confidence interval was set to 95%. Reported p-values were adjusted for multiple comparisons. Statistical analyses can be found in S1-S3 File.
Invasive growth patterns (plot profiles) were quantified as described [123]. Subtracting out the background signal and adjusting brightness and contrast were performed as stated above.
Images were imported into ImageJ and cropped based on invasive scar size (for example, SGAL: 200 x 200 pixels, SOE: 250 x 250 pixels, YPGAL: 300 x 300 pixels). A box was drawn across the midsection of the invasive scar, from edge-to-edge of the cropped image, with a pixel height of 40. Using the plot profile tool, which measures the grey value for pixels, a plot profile was generated for this region of the invasive scar. Plot profiles were generated along three different axes, which allowed an average value to be reported.
Invasive scar cross sections were performed by spotting equal concentrations of wild-type cells and the tec1Δ mutant onto SGAL medium. Plates were incubated for 7 d at 30˚C. Cells were washed off of plates in a stream of water. A razor blade and sterile forceps were used to excise the cells that had become embedded in the agar due to invasive growth. Chunks of agar containing invaded cells were placed on microscope slides to which a coverslip was added. For invasive squashes, gentle pressure was applied to squash the agar, revealing the trapped invaded cells. Cells were visualized by microscopy using the 5X, 10X, and the 100X objectives.

Measurement of MAPK pathway activity
To analyze MAPK pathway activity, the ß-galactosidase (lacZ) assay was performed. The following strains were used: wild type (PC313) and the tec1Δ (PC7675), dig1Δ (PC7676), opi1Δ (PC7674), rim101Δ (PC7673), and ras2Δ (PC6222) mutants. Cells were spotted on synthetic medium lacking uracil (SD-URA) to maintain selection for plasmids and were incubated for 21 h at 30˚. Cells were then scraped into dH 2 O and adjusted to the same optical density. Equal amounts of cells were then inoculated in indicated liquid media for 7 h, with shaking, at 30˚. Cells were harvested by centrifugation and stored at -80˚for at least 30 minutes. The ß-galactosidase (lacZ) assay was then performed as described [223,224] using a transcriptional reporter [pFRE-lacZ [217]] as the readout of MAPK pathway activity. The assay was performed with three biological replicates. In experiments that compare wild type across environments, values were normalized to YPD medium. In experiments that compare mutants to wild type, values were normalized to wild type. Error bars represent standard deviation between trials. Significance was determined by Student's t-test.

Quantitative reverse transcription PCR
Quantitative reverse transcription PCR (RT-qPCR) was used to measure the relative expression of the genes SFG1, FLO11, NFG1, RGD2, RPI1, and TIP1. The following strains were used when indicated: wild type (PC538) and the opi1Δ (PC2847), ste12Δ (PC539), tec1Δ (PC569), rim101Δ (PC2953), and ras2Δ (PC562) mutants. For experiments measuring the expression of SFG1 and FLO11, cells were spotted and grown on YPD or YPGAL semi-solid agar media for 2 d at 30˚before harvesting. For experiments measuring the expression of NFG1, RGD2, RPI1, and TIP1, cells were spotted and grown on YPD (High Glu) semi-solid agar medium for 2 d at 30˚or inoculated in 5 mL YPGAL liquid medium for 5 h, with shaking, at 30˚before harvesting. Cells were stored at -80˚before RNA extractions.
RNA extraction (hot-acid phenol-chloroform extractions), RNA purification [purified using a QIAGEN RNeasy Mini Kit (catalog # 74104), concentration and purity measured with NanoDrop 2000C (Thermo Fisher Scientific)], RNA stability determination (by agarose gel electrophoresis), cDNA generation [generated with iScript Reverse Transcriptase Supermix, Bio-Rad, catalog # 1708841], and RT-qPCR were performed as described in [123]. The Bio-Rad CFX384 Real Time System was used to perform RT-qPCR with iTaq Universal SYBR Green Supermix (Bio-Rad, catalog # 1725121). Primer sequences are shown in S4 Table. The 2 −ΔCt formula was used to calculate relative gene expression [138,225]. Ct was defined as the cycle where fluorescence was statistically significant above background. ΔCt is the difference in Ct between a target gene and the housekeeping gene ACT1. Three biological replicates were used, and average values were reported. Error bars represent standard deviation between trials. Significance was determined by Student's t-test.
The repository YEASTRACT [http://www.yeastract.com/index.php] was used to assess pathway-dependent changes in gene expression.

Cell adhesion measurements in liquid and semi-solid agar media
For adhesion assays, the following strains were used when indicated: wild type (PC538) and the opi1Δ (PC2847), tec1Δ (PC569), rim101Δ (PC2953), ras2Δ (PC562), dig1Δ (PC3039), and flo11Δ (PC1029) mutants. The adhesion of cells in liquid media was analyzed as described [123]. Cells were grown at 30˚for 24 h in 2 ml of specified medium with shaking. Images of groups (clusters) of cells were captured by microscopy at 5X magnification. Images were imported into ImageJ to subtract the background signal and apply a threshold to convert images to binary pixel images, where cells appeared black and the background appeared white. The same parameters were used for each image. The analyze particles tool was then used to measure the average area of cell clusters. The average of three biological replicates was determined and normalized to wild-type cells, which was set to 1. Error bars represent the standard deviation between trials. Significance was determined by Student's t-test.
Cell adhesion on agar media was quantified as described [123]. Cells were spotted onto semi-solid agar medium (as indicated) and incubated at 30˚for 7 d. Cells were harvested and resuspended in dH 2 O in 50 ml conical tubes. Tubes were inverted vigorously 10 times, and the contents of the tube were poured into a petri dish, where groups of adherent cells formed "adhesive particles" visible by eye. Particles were imaged by ChemiDoc XRS+ molecular imager under blot/chemicoloric setting with no filter. Images were digitally cropped by ImageJ using the circle tool around the edge of the Petri dish and background signal was subtracted. The same threshold was applied to each image to generate binary pixel images, where particles appeared black and the background appeared white. Using the analyze particles tool, total particle area was measured to represent total adhesion. Total adhesion was normalized to wild type, which was set to 1, and displayed as relative total adhesion. The average of three biological replicates was determined, where error bars represent the standard deviation between experiments. Significance was determined by Student's t-test.

Plastic adhesion assay
Measurement of the adhesion of cells to plastic was performed as described [134]. The following strains were used: wild type (PC538) and the opi1Δ (PC2847), tec1Δ (PC569), rim101Δ (PC2953), ras2Δ (PC562), dig1Δ (PC3039), and flo11Δ (PC1029) mutants. Cells were spotted onto semi-solid agar medium (as indicated) and incubated at 30˚for 7 d. Cells were removed from the community surface with a toothpick, resuspended in dH 2 O, and adjusted to the same optical density. Equal amounts of cells were added to polystyrene wells (Falcon Microtest Tissue culture plate, 96 Well) and incubated for 4 h at 30˚. Crystal violet dye (DIFCO) was then added for 20 min, and wells were washed equally and imaged with a Nikon D3000 digital camera. Quantification was performed by ImageJ analysis as described [123]. Each well was digitally cropped by the circle tool, equally from the center of the well. An equal threshold was used to convert images to binary pixel images, where adherent stained cells appear black against a white background. The analyze particle tool was used to measure the total area of adherent cells. Values were normalized to wild type that was set to 1. The experiment was performed in duplicate, and error was displayed as standard deviation from biological replicates.
Supporting information S1 Fig. Full survey of network mutants by the PWA. A) PWA on indicated media. First column, cells before wash, second column, inverted images of scars after wash, bar, 0.5 cm. Quantification in S2 Fig. B) PWA on +KCl medium; details in panel A. Only the dig1Δ mutant invades in this environment. C) All replicates on all media for wild type and the mutant strains were plotted for invasion versus colony size (mm 2 ). No correlation between invasion and colony size occurred (R 2 = 0.0004). (PDF) S2 Fig. Quantification of the PWA. A) PWA; Levels of relative invasion to wild type in indicated media, with wild-type values set to 1. Asterisk, p-value � 0.05, compared to wild type. (Images in S1A Fig) B) Levels of relative invasion for the dig1Δ mutant, with wild-type values set to 1; Asterisk, p-value � 0.05, compared to wild type by Student's t-test. (PDF)

S3 Fig. Invasive growth network ranking. A)
Ranking of network components per their regulatory role of invasive growth in environments where invasive growth occurs (11 out of 12 environments, excludes +KCl). A) The average relative invasion compared to wild type (set to 1) was calculated across all environments for network mutants. Error represents standard deviation. B) Total number of environments a network mutant met a threshold of decreased invasion relative to wild type for indicated thresholds: � 75%, � 50%, � 25%, � 20%, � 15%, or � 10% of wild-type invasion. Each threshold was totaled independently. The pathways were ranked in order using all thresholds (# of thresholds). (PDF)

S4 Fig. Growth of wild-type cells and mutants lacking the major regulatory pathways that control invasive growth.
Wild-type cells and the indicated mutants were grown in the indicated media in liquid culture for 16h at 30˚by shaking. Cells were washed once in water, and growth was measured by OD at 600nm. For each condition, wild-type values were set to a value of 1. The average of three replicates is reported. Error is reported as standard deviation. Top row, cells before wash, bottom row, inverted images of scars after wash, bar, 0.5 cm. B) Same as panel A, except on SD medium. C) We found a strong temporal role on SD medium but not YPGAL medium for the dig1Δ mutant. Levels of relative invasion between wild type and the dig1Δ mutant on indicated medium, with wild type values set to 1. Left, YPGAL, right SD. Asterisk, P-value � 0.05, compared to wild type. D) Levels of relative invasive growth versus relative MAPK pathway activity in wild-type cells. E) Levels of invasive growth on YPGAL or SD media over 21 d for wild type (black) and the dig1Δ (purple) and tec1Δ (green) mutants. (PDF) The more ruffly a complex-colony morphology the more adhesion between cells. Images of colony after 7 d of growth on indicated media, bar, 0.5 cm. Wild-type cells show strong ruffling on YPD and YPGAL media and weaker but still increased complexcolony morphology compared to the flo11Δ mutant on SOE and SD media. The flo11Δ mutant exhibited a smooth pattern in all environments. All mutants tested showed reduced complexcolony morphology compared to wild type. Compared to the flo11Δ mutant the following mutants showed increased complex-colony morphology on indicated media: tec1Δ mutant on YPGAL medium; ras2Δ mutant on YPD medium; rim101Δ mutant on SOE medium; opi1Δ mutant on YPD medium; dig1Δ mutant on all media. B) Total cell adhesion within a colony. Images of adherent cells from the colony surface are seen as black particles. Bar graphs, levels of relative total adhesion compared to wild type, with wild-type values set to 1. Asterisk, Pvalue � 0.05, compared to wild type. The flo11Δ mutant exhibited no detectable adhesion within the colony on any environment. Wild-type cells showed cell-cell adhesion on all four environments. The tec1Δ and ras2Δ mutants showed increased adhesion on YPGAL medium compared to the flo11Δ mutant. (PDF) S12 Fig. Plastic adhesion assay. Plastic adhesion is a medically relevant phenotype because pathogenic yeasts, like Candida albicans, will adhere to medical devices and plastics in hospital settings. Images of stained cells adhering to a polystyrene plastic 96-well plate. Bar graph, quantification of relative plastic adhesion to wild type, with wild type values set to 1. (PDF) S13 Fig. MAPK regulates target genes in one environment but not another. Targets of the MAPK pathway, NFG1, RGD2, RPI1, and TIP1 identified previously [100], were regulated by MAPK in one environment but not another. A) RT-qPCR analysis of mRNA levels for indicated genes between wild-type and the ste12Δ mutant (a MAPK pathway mutant equivalent to tec1Δ) in YPGAL medium. Wild-type values were normalized to ACT1 expression and set to 1. Asterisk, p-value � 0.008, compared to wild type. RT-qPCR data for YPGAL comes from [100]. B) Same as panel A, except on YPD (High Glu) medium. C) These genes also show different changes in expression between the two environments. RT-qPCR analysis of mRNA levels between YPGAL and YPD (High Glu). Wild-type values were normalized to ACT1 expression. YPD (High Glu) values were set to 1. Asterisk, p-value � 0.005, comparing YPGAL to YPD (High Glu). (PDF)