You and I: The Effects of Intra- and Inter-Species Interactions on Microbial ,

Growth, and Morphology

by

Christopher James Cody Davis

Department of Duke University

Date:______Approved:

______Paul Magwene, Supervisor

______Rytas Vilgalys

______Fred Dietrich

Thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in the Department of Biology in the Graduate School of Duke University

2016

i

v

ABSTRACT

You and I: The Effects of Intra- and Inter-Species Interactions on Microbial Biofilms,

Growth, and Morphology

by

Christopher James Cody Davis

Department of Biology Duke University

Date:______Approved:

______Paul Magwene, Supervisor

______Rytas Vilgalys

______Fred Dietrich

An abstract of a thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in the Department of Biology in the Graduate School of Duke University

2016

i

v

Copyright by Christopher James Cody Davis 2016

Abstract

Humanity is shaped by its relationships with microbes. From bacterial infections to the production of biofuels, industry and health often hinge on our control of microbial populations. Understanding the physiological and genetic basis of their behaviors is therefore of the highest importance. To this end I have investigated the genetic basis of plastic adhesion in Saccharomyces cerevisiae, the mechanistic and evolutionary dynamics of mixed species biofilms with and S. cerevisiae, and the induction of filamentation in E. coli. Using a bulk segregant analysis on experimentally evolved populations, I detected 28 genes that are likely to mediate plastic adhesion in S. cerevisiae. With a variety of imaging and culture manipulation techniques, I found that particular strains of E. coli are capable of inducing flocculation and macroscopic formation via coaggregation with yeast. I also employed and microbial demography techniques to find that selection for mixed species biofilm association leads to lower fecundity in S. cerevisiae. Using culture manipulation and imaging techniques, I also found that E. coli are capable of inducing a filamentous phenotype with a secreted signal that has many of the qualities of a quorum sensing molecule.

iv

Contents

Abstract ...... iv

List of Tables ...... viii

List of Figures ...... ix

List of Abbreviations ...... xi

1. Plastic adhesion in S. cerevisiae ...... 1

1.1 Introduction ...... 1

1.2 Methods ...... 3

1.2.1 Strains ...... 3

1.2.2. Selection regime ...... 4

1.2.3 Detection of QTLs ...... 5

1.2.4 Quantification of plastic adhesion ...... 6

1.2.5 Measuring the effects of nutrient limitation on adhesion ...... 7

1.3 Results ...... 8

1.3.1 Potential QTLs are abundant ...... 8

1.3.2 Selection increased plastic adherence ...... 11

1.3.3 Nitrogen limitation increases adhesiveness ...... 15

1.4 Discussion ...... 16

2. Mixed species biofilms with S. cerevisiae and E. coli ...... 21

2.2 Introduction ...... 21

2.2 Methods ...... 23

v

2.2.1. Flocculation measurements ...... 23

2.2.2 Biofilms ...... 24

2.2.3 Long term co-culturing ...... 26

2.2.4 TrackScar ...... 28

2.3 Results ...... 28

2.3.1 DH5α and MACH1 induce flocculation via coaggregation with S. cerevisiae .. 28

2.3.2 DH5α and MACH1 form macroscopic biofilms when co-cultured with yeast 29

2.3.3 Biofilms are an intimate association of both species ...... 34

2.3.4 Biofilm formation requires live E. coli and is affected by deflocculation buffer ...... 37

2.3.5 Selection for biofilm association in S. cerevisiae leads to slower growth rates .. 38

2.4 Discussion ...... 40

3. Filamentation in E. coli ...... 45

3.1 Introduction ...... 45

3.2 Methods ...... 47

3.2.1 Culturing ...... 47

3.2.2 Imaging and analysis ...... 47

3.3 Results ...... 48

3.3.1 E. coli can produce a signal that induces their own filamentation ...... 48

3.3.2 There is variation in filamentation behavior among strains ...... 50

3.3.3 Strain variation in filamentation is due to differences in reaction to inducing agent, not production of it ...... 51

3.3.4 The PFIA is heat stable up to 80°C ...... 52

vi

3.3.5 Glucose is necessary for the production of the PFIA ...... 53

3.3.6 Low pH conditions and acclimatization are necessary for filamentation to occur ...... 54

3.4 Discussion ...... 56

4. Conclusions ...... 61

References ...... 63

vii

List of Tables

Table 1: List of genes implicated in plastic adhesion by BSA/PROVEAN pipeline……...10

viii

List of Figures

Figure 1: G-statistic plots for PMY1683 and PMY1846...... 9

Figure 2: Plastic adherence over the course of the selection regime in PMY1643, PMY1683, PMY1846, and PMY1847...... 13

Figure 3: Colony counts over the course of the selection regime in PMY1643, PMY1683, PMY1846, and PMY1847 ...... 14

Figure 4: Distribution of all adherence assay measurements...... 15

Figure 5: Effects of nitrogen and glucose limitation on plastic adhesion ...... 16

Figure 6: Effects of coculture with DH5α E. coli on floc size in a panel of S. cerevisiae strains ...... 30

Figure 7: Effects of coculture with DH5α on floc size in a panel of S. cerevisiae FLO knockout strains ...... 31

Figure 8: Effect of DH5α filtrate on floc size in PMY127 S. cerevisiae...... 32

Figure 9: Image of a large floc formed by fluorescent strains of MACH1 E. coli and PMY1691 S. cerevisiae...... 32

Figure 10: Photos of co-culture dependent biofilms with DH5α and MACH1 ...... 33

Figure 11: Co-culture dependent biofilms with a variety of strains ...... 33

Figure 12: Images of biofilms formed by fluorescently tagged strains of MACH1 and PMY1690 S. cerevisiae...... 35

Figure 13: A time course of biofilm formation in co-cultures of fluorescently tagged MACH1 and PMY1690 ...... 36

Figure 14: Effects of deflocculaton buffer on the structure of co-culture biofilms...... 37

Figure 15: Maximum growth rates in YPD of CMY98 S. cerevisiae strains from co-culture selection experiment ...... 38

ix

Figure 16: Averages and distributions of individual fecundity in biofilm-selected and non-selected populations of CMY98 ...... 39

Figure 17: A comparison of normal and filamentous MACH1 cells ...... 49

Figure 18: ASLDs of MACH1 after overnight culture in a variety of media conditions .. 50

Figure 19: ASLDs of five E. coli strains when cultured in YPD and their own filtrates .... 51

Figure 20: MACH1 and HB101 E. coli ASLDs when cultured in one another’s filtrates ... 52

Figure 21: ASLDs of MACH1 grown in heat-treated filtrate ...... 53

Figure 22: ASLDs of MACH1 in LB, LB 2% glucose, and YPD media and filtrates ...... 54

Figure 23: ASLDs of MACH1 filtrate treatments with variable pH at time of inoculation ...... 55

Figure 24: ASLDs of MACH1 filtrate treatments with inoculums from LB or YPD cultures ...... 55

x

List of Abbreviations

AHL – acyl-homoserine lactone

AI-2 – autoinducer-2

ANOVA – analysis of variance

ASLD – average size of the largest decile of cells cAMP – cyclic adenosine monophosphate

CAP – catabolite activator protein

BSA – bulk segregant analysis

DPD – 4,5-dihydroxy- 2,3-pentanedione

GPI – glycosylphosphatidylinositol

LB – lysogeny broth

PFIA – putative filamentation-inducing agent

QTLs – quantitative trait loci

SC – synthetic complete

YPD – yeast extract peptone dextrose

xi

1. Plastic adhesion in S. cerevisiae

1.1 Introduction

The ability of yeast to adhere to abiotic surfaces has likely been observed ever since the advent of brewing some 7000 years ago (Mortimer 2000). By comparison, the invention and proliferation of plastics is an extremely recent development. Recently created or not, abhors an uncolonized substrate, and numerous species of microbes have shown the ability to adhere to plastics. Fungal adhesion to plastics is of particular interest to the medical community, as biofilms adhering to catheters and prostheses can cause disseminated fungal infections with mortality rates exceeding 40%

(Wisplinghoff et al. 2004). Biofilm formation on plastics is also a relevant issue for the brewing and biofuel industries, where accumulations of yeast on equipment can affect production processes. A greater understanding of fungal plastic adhesion would therefore be widely useful, and the extensive molecular and genetic resources available for S. cerevisiae make it a premier model organism to study it in.

The main players in fungal adhesion and flocculation are cell wall proteins called adhesins (K. J. Verstrepen and Klis 2006). Different S. cerevisiae adhesins allow for adhesion to different surfaces, but all share the same basic structure. At one end there is a glycosylphosphatidylinositol (GPI) anchor attaching the protein to the cell wall (Bony et al. 1997; Kapteyn, Van Den Ende, and Klis 1999). At the other end there is generally a peptide or carbohydrate binding domain (Kobayashi et al. 1998; Rigden, Mello, and

1

Galperin 2004). The majority of the protein is made up of the middle section, a chain of threonine and serine rich repeats. Variation in the number and order of these repetitive sequences is thought to account for the diversity of binding affinities found among different adhesins with essentially the same structure (K. J. Verstrepen and Klis 2006)

The primary adhesin implicated in S. cerevisiae’s adhesion to plastics is Flo11p, which increases the hydrophobicity of the cell surface, allowing it to adhere to certain substrates via hydrophobic interactions (Guo et al. 2000; Kang and Choi 2005; Reynolds and Fink 2001). The regulatory pathways controlling Flo11p production are numerous and well-studied, with the MAPK, Ras-cAMP, TOR, and main glucose repression pathways all being implicated, but knowledge of the upstream sensors affecting these pathways is relatively sparse (Madhani and Fink 1997; Rupp et al. 1999; Gagiano et al.

1999; Vyas et al. 2003; Schwartz and Madhani 2004; K. J. Verstrepen and Klis 2006).

Flo11p is not the only protein at play in plastic adhesion. Although the mechanism is unknown, glucose and nitrogen starvation have been shown to induce increased cell surface hydrophobicity and plastic adhesion in a Flo11p-deficient strain (Mortensen et al. 2007). The genetic factors contributing directly and indirectly to plastic adhesion are therefore incompletely understood.

In order to identify quantitative trait loci (QTLs) that contribute to plastic adhesion I conducted a bulk segregant analysis (BSA) on populations of S. cerevisiae that had been selected for increased adhesion ability. I used two clinical isolate to found

2

populations, selected for plastic adhesion, and compared the genomes of the highly adhering selected populations to the weakly adhering non-selected populations. From these comparisons genomic regions containing likely QTLs were identified. I then analyzed the polymorphisms in the genes contained in these regions to identify individual genes likely to affect plastic adhesion. Additionally, I carried out assays to quantify the strength of response to the selection regime and the effects of nutrient limitation on plastic adhesion.

1.2 Methods

1.2.1 Strains

Four homothallic, haploid strains of S. cerevisiae were used. Two strains were derived from the parent strain PMY144, a clinical isolate. One, PMY1847, was tagged with mCherry on the metabolic protein PGK1 and cultured under selection of the G418 to preserve this tag. The other, PMY1846, was not. The other two strains were derived from parent strain PMY127, also a clinical isolate. One, PMY1643, was the product of two successive rounds of sporulation (F2 generation). The other, PMY1683, was the product of eight successive rounds of sporulation (F8 generation).

Prior to the selection regime, PMY1846 and PMY1847 were sporulated and germinated to produce populations of haploid segregants. PMY1643 and PMY1683 were already comprised of haploid segregants and were grown from frozen stock overnight.

After overnight growth at 30° C in 5 mL yeast extract peptone dextrose (YPD; 1% yeast

3

extract, 2% peptone, 2% glucose), 50 µL of each strain’s culture was used for the initial inoculation of experimental and control treatments.

1.2.2. Selection regime

Each strain underwent a selection regime consisting of three treatments: the sonication treatment, the bead-to-bead transfer treatment, and the control treatment. In each treatment strains were grown in 15 mL glass tubes containing 5 mL of chemically defined low glucose synthetic complete (SC; 0.67% yeast nitrogen base, 0.2% amino acids, 0.5% glucose) media. Transfers to new tubes were made every 48 hours for 10 cycles. Immediately before each transfer a 900 µL aliquot of culture was taken, combined with 900 µL of a 30% glycerol solution, and frozen at -80° C.

In the sonication treatment a single sterile polystyrene bead was placed in the tube at the time of inoculation. At transfers the bead was isolated with sterile forceps and washed in 2 mL of water in the well of a 24 well plate. After washing, the bead was placed in a 15 mL plastic culture tube containing 3 mL of sterile water and sonicated with three 1-second long bursts using a Branson Sonifier 150 on power setting 4. The bead was discarded after sonication, and a 30 µL aliquot of the sonicate was spread onto a YPD plate and incubated overnight at 30°C. Colonies were counted as a measure of the amount of cells sonicated off the bead. In treatments where 30 µL of sonicate resulted in growth too dense to distinguish individual colonies, smaller amounts of sonicate were diluted to 30 µL and plated. After the plating aliquot was taken, the remaining 2.97 mL

4

of sonicate was centrifuged and resuspended in 200 µL of water, then used to inoculate a new tube of SC media with a new bead. Three replicates of the sonication treatment were maintained throughout the experiment. In the bead-to-bead transfer treatment, two beads were placed in the tube. One bead was marked with a sharpie and the other was not. At the time of transfer one bead was extracted, washed, and put into a fresh SC tube along with an oppositely marked bead. At the next transfer the newer bead was transferred and the older one was discarded. In the case of PMY1847, the surface of the older bead was imaged using fluorescent microscopy before it was discarded. The bead- to-bead transfer treatment also had three replicates. In the control treatment transfers were made by taking 50 µL of culture and directly inoculating a new SC tube with it.

Controls had two replicates. To get control measurements of bead adhesion, control populations from three transfers were put through the same procedures used in the sonication treatment.

1.2.3 Detection of QTLs

After selection was completed, DNA libraries were constructed from the sonication treatment and control treatment populations. Genomic DNA extractions were done using a Quiagen genomic-tip DNA isolation kit, fragmentation was done with a

NOT1 digest, and DNA libraries were constructed using the NEBNext Ultra DNA library prep kit for Illumina. Currently, only libraries for the sonication and control treatments of PMY1846 after the third transfer and PMY1683 before the first transfer

5

have been sequenced. Sequencing was done on the Illumina HiSeq platform using single end reads. Reads were then aligned to a reference genome and BSA G-statistics were generated using a script written by Paul Magwene (Magwene, Willis, and Kelly 2011).

Once genomic regions containing large G-statistic peaks (>10) were identified, they were inputted into a custom pipeline. This pipeline took the constructed high and low bulk genomes and compared their sequences for every gene documented in the

Saccharomyces Genome Database contained under the peaks (Cherry et al. 2012). If a polymorphism was detected, the sequences of the two bulk’s alleles were translated into peptides and inputted to a program called PROVEAN (Choi 2012). PROVEAN is a program that estimates the likelihood that a difference in the primary structure between two protein alleles will cause a difference in function. The threshold for a significant

PROVEAN score was set at -0.5. This threshold was set using FLO11’s PROVEAN score from the PMY1846 comparison, the lowest scoring residue of which was -0.813. Genes containing residues that were estimated by PROVEAN to cause a difference in function between high and low bulks were compiled into a list for further investigation.

1.2.4 Quantification of plastic adhesion

An imaging-based assay was used to quantify the plastic adhesion strength of populations. First, frozen stock of a population was streaked onto a YPD plate and grown overnight at 30° C. Individual colonies were then isolated and transferred into the wells of a round bottom 96 well plate containing 200 µL of YPD per well and grown

6

overnight. To help standardize the amount of cells in the final inoculum, the next day cells were pinned into another round bottom 96 well YPD plate and grown overnight.

The next day the cells were resuspended by pipetting and 5 µL of culture from each well was used to inoculate the wells of a flat bottom 96 well plate containing 200 µL of low glucose SC media per well. After a 36-hour incubation at 30° C, the media in these plates was poured out and the plates were submerged in deionized water three times. After drying for two hours, the plates were imaged at 5X magnification using a Zeiss Axio

Observer microscope. After images were obtained, a Fiji script written by Daniel Skelly was used to calculate the percentage of the well bottom covered by adhering yeast cells

(Schindelin et al. 2012). Adhesiveness of a population was quantified as the average percentage of the well bottom covered by its segregants’ cells. All tests for statistical significance were done using one-way ANOVAs in R and all plots were made using the ggplot2 package (Wickham 2009).

1.2.5 Measuring the effects of nutrient limitation on adhesion

To measure the effects of carbon and nitrogen limitation on plastic adhesion, a flat bottom 96 well plate containing 2:1 serial dilutions of different nutrient and media types was made. Three media were tested: SC, YPD, and lysogeny broth (LB; 0.5% yeast extract, 1% tryptone, 1% NaCl), all containing 2% glucose. For each media type, four dilution columns were made: two where nitrogen base (yeast nitrogen base for SC, peptone for YPD, and tryptone for LB) was diluted and two where glucose was diluted.

7

In the nitrogen dilution columns, nitrogen varied from 100% of the media’s normal value to 1.563%, including a 0% control, while glucose remained at 100% in all wells.

The glucose dilutions were the inverse. In one of the two dilution columns of a given media and nutrient, a segregant of PMY1683 chosen for its medium strength adhesion

(approximately 75% in previous adhesion assays) was cultured. In the other column of the same media and nutrient dilution, the DH5α strain of E. coli was cultured. Each well contained 200 µL of media and was inoculated with 5 µL of culture. After 24 hours of growth at 30° C in a Tecan Sunrise microplate reader, the plate was poured out, repeatedly dunked in water, and dried as descried in the adhesion assay protocol.

Imaging and analysis was also done using the methods previously described for the adhesion protocol.

1.3 Results

1.3.1 Potential QTLs are abundant

G-statistic plots show no defined peaks for the comparison between the starting populations for the experimental and control bulks of strain PMY1683 (Figure 1a). At this point no selection events had happened, so no differences in allele distribution between the two populations would be expected. In contrast, the G-statistic plot from the bulks of PMY1846 after three selection events showed several distinguishable peaks

(Figure 1b).

8

Figure 1: a. A map of G-statistic peaks across the genome from a BSA of PMY1683 experimental and control strains before any selection events have caused them to diverge. b. A G-statistic map of PMY1846 experimental and control strains after three selection events.

28 genes fell under G-statistic peaks and contained at least one polymorphism with a PROVEAN score less than -0.5. (Table 1). These genes are the most promising candidates for influencing plastic adhesion to come out of this analysis. Their functions are diverse. Two adhesin genes, FLO11 and FIG2, correspond to the tallest G-statistic peaks in the analysis, located on chromosomes nine and three, respectively. These two also contain the most polymorphisms of any in the set, with 13 for FLO11 and 18 for

9

FIG2. Other genes with more than 5 polymorphisms include EDS1 (8), SEG2 (11), HSL1

(7), and MSB1 (6). Unlike FLO11 and FIG2, there is no obvious functional similarity connecting these genes.

Table 1: A list of the 28 most likely gene candidates from the BSA/PROVEAN pipeline and their putative functions. All functions are from the Saccharomyces Genome Database (Cherry et al., 2012).

Involved in formation of early endocytic sites, binds membranes and EDE1 possibly membrane bound proteins PSY4 Regulatory subunit of protein phosphotase PP4 FLR1 Plasma membrane transporter of the major facilitator superfamily CSG2 Endoplasmic reticulum membrane protein EDS1 Putative zinc cluster protein, predicted to be transcription factor FMP23 Putative protein of unknown function FIG2 Cell wall adhesin expressed specifically during mating YCL049C Unknown function, localizes to membrane fraction SRO9 Cytoplasmic RNA-binding protein ABP1 Actin-binding protein of the cortical actin cytoskeleton GPI-anchored cell surface glycoprotein (flocculin); required for FLO11 pseudohyphal and invasive growth, flocculation, and biofilm formation Plays a role in phosphatidylinositol 4,5-bisphosphate homeostasis and in INP51 endocytosis DJP1 Required for peroxisomal protein import Integral plasma membrane protein; involved in the synthesis of the LAS21 glycosylphosphatidylinositol (GPI) core structure ABF1 DNA binding protein with possible chromatin-reorganizing activity YKL107W Putative short-chain dehydrogenase/reductase SEG2 Eisosome component APE1 Vacuolar aminopeptidase HSL1 Nim1p-related protein kinase Regulates abundance of high-affinity plasma membrane transporters YPF1 during the starvation response CWP1 Cell wall mannoprotein that localizes to birth scars of daughter cells Cell wall mannoprotein; plays a role in maintenance of newly CCW12 synthesized areas of cell wall LCL2 Putative protein of unknown function 10

LAM6 Sterol transporter that transfers sterols between membranes Protein of unknown function; may be involved in positive regulation of MSB1 1,3-beta-glucan synthesis and the Pkc1p-MAPK pathway GIP3 Cytoplasmic protein that regulates protein phosphatase 1 Glc7p GPH1 Glycogen phosphorylase required for the mobilization of glycogen PSE1 Karyopherin/importin that interacts with the nuclear pore complex

1.3.2 Selection increased plastic adherence

Plastic adherence assays of strains PMY1643 and PMY1683 show that the selection regime increased adherence significantly (p<.01) and that most gains in adherence occurred after the first selection event (Figures 2a and 2b). In contrast, selected populations of PMY1846 and PMY1847 did not show any increase over controls or their own starting populations (Figures 2c and 2d). No samples from the PMY144 background ever had greater than 4% coverage on the adherence assay (data not shown). However, colony counts from bead sonicates indicate that the number of cells adhering to the bead increased over time in the PMY144 background strains (Figure 3a), although the increase was dwarfed by that of the PMY127 background strains (Figure

3b). This would suggest that the PMY144 background strains did increase in adherence over the course of selection, but not enough to be detectable by the adherence assay. The distribution of all samples from all adherence assays shows bimodality reflecting the radically different adherence profiles of the two parent strains (Figure 4).

11

12

Figure 2: a. Changes in average plastic adherence in the PMY1643 sonication and control treatments over the course of the selection regime b. PMY1683 sonication and control treatments c. PMY1846 sonication and control treatments – note the much smaller y-axis values d. PMY1847 sonication and control treatments.

13

Figure 3: a. Colony counts of PMY144 derived strains’ sonication and control treatments over the course of the selection regime. b. PMY127 derived strains’ colony counts. Note the roughly order of magnitude increase compared to the PMY144 strains.

14

Figure 4: Density distribution plot of all adherence assay measurements made. Strains derived from the weakly-adhering PMY144 make up the left peak of the distribution, while strains from the strongly-adhering PMY127 make up the right side.

1.3.3 Nitrogen limitation increases adhesiveness

In the nutrient starvation adhesion assay, only the S. cerevisiae SC media treatments had any wells score greater than 30% cell coverage (data not shown). In the

SC treatments, a spike in adherence was seen when nitrogen base concentrations dropped below 0.42 g/L. Some adhesion was still seen at 0 g/L, likely due to cells using the amino acids included in SC as a nitrogen source (Figure 5a). No increase in adherence was seen in low glucose conditions (Figure 5b).

15

Figure 5: a. Plastic adhesion assay scores of a PMY1683 segregant as a function of nitrogen availability in SC media. b. Adhesion assay scores for the same segregant as a function of glucose availability.

1.4 Discussion

The candidate genes revealed by the BSA and PROVEAN analysis are diverse in

16

function. Some, like FLO11, are obvious targets for adhesion selection. FLO11 plays a major role in plastic adhesion, filamentation, biofilm formation, and invasive growth

(Bayly et al. 2005). It is also highly polymorphic, with 13 alleles found in a survey of 20 strains. These alleles vary primarily in the number of 36 bp long tandem repeats that make up the middle of the gene (Zara et al. 2009). The length of FLO11 alleles has been found to influence various phenotypes, and it is likely that the regime here selected for different length alleles between the two bulks (Fidalgo, Barrales, and Jimenez 2009). The high number of polymorphic sites detected between the high and low bulk FLO11 alleles could be explained by the large indels that would result from such selection. FIG2 is another cell wall adhesin gene detected by our analysis. FIG2 is canonically involved in cell-cell adherence during mating, but FIG2 overexpression can restore filamentation in FLO11 knockouts, and FLO11 overexpression can restore mating adherence in FIG2 knockouts (Guo et al. 2000). In light of this functional overlap, it is not surprising that both genes could have an effect on plastic adhesion. LAS21 is involved in the synthesis of glycosylphosphatidylinositol (GPI) (Benachour et al. 1999). Flo11p and other adhesins are attached to the cell wall with GPI anchors, providing a clear reason why LAS21 would be affected by the selection regime (K. J. Verstrepen and Klis 2006).

The role of other candidate genes in plastic adhesion are less obvious. Several are regulatory in nature, making enticing targets for future studies on the regulatory pathways that control adhesion. Two, CWP1 and CCW12, code for cell wall

17

mannoproteins. While it is safe to say they affect the qualities of the cell wall, it is unclear how those effects are involved in plastic adhesion. EDE1, SEG2, and INP51 are all surprisingly involved in endocytosis. Their presence could suggest that the cellular machinery involved in endocytosis somehow affects plastic adhesion, but could also be the result of unintentional pressures from the selection regime. For instance, if the microenvironment of the bead surface was more crowded than the surrounding media, it may have created increased competition for suspended nutrients and thus selection for different alleles of endocytosis genes. Going forward, more populations from the experiment should be sequenced and analyzed. It is especially important that the highly adhering PMY127-derived strains be analyzed, as evidence for the successful creation of high- and low-adhering bulks is much stronger for them than the PMY144-derived strains. Genes that are repeatedly identified as differing between the experimental and control populations should then be analyzed with a battery of knockout and overexpression experiments.

Plastic adherence assays on PMY127-derived strains PMY1643 and PMY1683 show that the selection regime was effective in increasing the average adherence of experimental populations. In contrast, PMY144-derived strains PMY1846 and PMY1847 showed no detectable increase in adhesiveness. However, colony counts from sonicate plating indicated an increase in the number of cells colonizing the bead over time in the

PMY144 background strains. This suggests that while the selection regime did increase

18

the average adhesiveness of the PMY144 strains, it did so at such a small magnitude it was undetectable by the plastic adherence assay. Evolutionary responses are limited by the genetic resources available, and it would appear that PMY144 is severely lacking in pro-adhesion alleles compared to PMY127. The failure of PMY144-derived experimental populations to score higher in the adherence assays should not be interpreted as a failure of the selection regime to select for the desired phenotype, but rather as a failure of the assay to accurately distinguish differences at very low levels of adherence. In addition to the colony count data, the presence of FLO11 in the candidate gene list further supports the position that the selected PMY144 populations represent a legitimate high bulk for BSA analysis and that the candidate genes detected from them are appropriate targets for further analysis. However, BSA analysis should be done on the PMY127 bulks and candidate genes should be cross verified between the two backgrounds before gene manipulation experiments are undertaken.

Carbon and nitrogen limitation have both been demonstrated to induce flocculation, FLO11 expression, and invasive growth in S. cerevisiae (Smit et al. 1992; K.

Verstrepen et al. 2003; Bayly et al. 2005; Lo and Dranginis 1998; Cullen and Sprague

2000). It is therefore surprising that only nitrogen limitation was found to increase plastic adherence in PMY1683. This result suggests that the response pathways to these two types of nutrient starvation lead to different adherence phenotypes. However, it is unclear whether this result can be generalized to all S. cerevisiae strains, or if PMY1683 is

19

unique in its reaction to nitrogen and glucose limitation. Furthermore, these findings are based off a single experiment and must be replicated before any conclusions about even

PMY1683’s behaviors are drawn. If further experimentation shows that this reaction profile is generalizable to many strains, expression assays should be used to dissect its genetic basis.

20

2. Mixed species biofilms with S. cerevisiae and E. coli

2.2 Introduction

Microbes in nature primarily exist in multicellular aggregates called biofilms

(Granek et al. 2013). Biofilms impact humanity in a number of ways. Industrially, they can be employed in wastewater treatment and bioremediation, but can also clog pipes and corrode surfaces (Dunne 2002). Medically, they can be sources of chronic infection, but can also protect against colonization by pathogenic species (Dunne 2002; Vlamakis and Kolter 2010). Due to their economic and medical relevance, biofilms have been heavily studied, primarily using species in isolation. However, the majority of biofilms in nature consist of a complex assemblage of species that represent distinct ecosystems

(Kolter and Greenberg 2006). More recently much interest has been shown in mixed species biofilms, which can facilitate symbiotic interactions such as cross-feeding and resistance to antimicrobial agents that neither species can achieve in monoculture

(Wolcott et al. 2013; Adam, Baillie, and Douglas 2002). Due to the diverse makeup of biofilms in nature and especially the human body, such interactions demand further investigation.

This investigation began with a simple observation: in cultures containing both S. cerevisiae and E. coli cells, flocs were larger than in cultures containing only S. cerevisiae cells. Shortly after this discovery it was found that overnight co-cultures also formed

21

thick biofilms near the liquid-air interface. Confocal microscopy showed that both species were present in the interface biofilm, making it a potentially fruitful system for investigating mixed species biofilms from both a mechanistic and evolutionary perspective.

Mechanistic studies began by finding the specific conditions necessary for biofilm formation. All yeast strains and species tested formed biofilms in co-culture, but biofilm formation was specific to only two E. coli strains. These two strains were DH5α and MACH1, both lab strains commonly used for cloning. The two strains are quite diverse – DH5α is derived from parent strain K-12, isolated from the stool of a diphtheria patient at in 1922 (Bachman 1972). MACH1 is derived from parent strain W, isolated from the soil of a cemetery near Rutgers University in

1943 by (Archer et al. 2011). The same behavior being displayed by strains separated by 20 years, 3000 miles, and radically different environments suggested that it is not simply the result of a one-off mutation.

Experiments with have shown that the evolutionary trajectories and resource usage profiles of species in polyculture are different than those in monoculture

(Lawrence et al. 2012). It has also been shown that the bacteria Staphylococcus hominis can induce S. cerevisiae to use alternative carbon sources in the presence of glucose (Jarosz et al. 2014). To investigate the evolutionary and metabolic effects of co-culture and biofilm association on S. cerevisiae and E. coli, I imposed an artificial selection regime via long

22

term co-culture and measured its effects on resource usage and fecundity.

2.2 Methods

2.2.1. Flocculation measurements

Floc measurement experiments were conducted by first growing E. coli and S. cerevisiae separately in 15 mL plastic culture tubes containing 5 mL YPD overnight at 30

°C on a rolling incubator. All cultures mentioned here were grown under the same conditions unless otherwise noted. Experimental tubes were made by putting 25 µL of each species’ overnight culture into 5 mL of YPD, while control tubes were made by putting 25 µL of only S. cerevisiae or E. coli into 5 mL YPD. The DH5α E. coli strain was used in all experiments presented except for the creation of fluorescent flocs, where

MACH1 was used. A 12 strain yeast panel consisting of the 11 most diverse yeast strains from the 100 genomes project plus PMY127 were used in the initial flocculation experiment (Strope et al. 2015). A panel of FLO knockouts from the Σ1278B and S288C backgrounds were also tested against DH5α.

An experiment on the effects of DH5α culture filtrate on PMY127’s floc size was also carried out. DH5α was grown in YPD for approximately 20 hours, then filtered through a 0.2 µm syringe filter. 2 mL of YPD was added to 3 mL of filtrate to ensure that sufficient nutrients were present for PMY127’s growth. After this mixture was prepared, it was inoculated with 25 µL of overnight PMY127 culture.

After overnight incubation, 2 µL of each culture were drawn from the media and

23

imaged using an AMG Evos microscope. Five images of each treatment were taken at

20X magnification. In between images, the screen was turned off and the stand was moved at random in order to avoid bias towards imaging larger flocs. Once images were taken, polygons were drawn around the boundaries of flocs and their areas were calculated using Fiji software (Schindelin et al. 2012). Any group of five or more cells touching was considered a floc. As in all the following, statistical tests for significance were done using one-way ANOVAs in R and plots were made using the ggplot2 package (Wickham 2009).

2.2.2 Biofilms

Standard mixed species biofilms were formed using the following procedure: overnight monocultures of E. coli and S. cerevisiae were grown in plastic 15 mL cell culture tubes containing 5 mL YPD, then 25 µL of both species’ cultures were put into another plastic 15 mL tube with 5 mL of YPD and co-cultured overnight. When growing biofilms for macroscopic imaging, the same procedure was used except that culture tubes were 5 mL, contained 2 mL YPD, and were seeded with 10 µL of each species’ culture. The presence of a macroscopic interface biofilm was ascertained through simple visual inspection after approximately 18 hours of growth.

In heat killing experiments, 5 mL overnight YPD cultures of either E. coli or S. cerevisiae were placed in a hot water bath at 80° C for one hour. DH5α was the only E. coli strain used, and the S. cerevisiae counterpart consisted of a panel of the PMY127,

24

PMY1504, PMY1535, and PMY1537 strains. To make sure heat killing was effective, a 50

µL aliquot was taken from each tube and used to inoculate 5 mL of YPD. No growth ever occurred in tubes inoculated from heat treated cultures, so heat killing was presumed to be 100% successful. After heat treating, 2 mL of the cultures were removed and replaced with 2 mL of fresh YPD to ensure sufficient nutrients for microbial growth.

Then 100 µL of the counterpart species’ overnight culture was used to inoculate the tube.

Macroscopic biofilm formation was ascertained through visual inspection after approximately 18 hours of incubation.

Filtration experiments used the same sets of strains as the heat killing experiments. Overnight 5 mL YPD cultures of E. coli or S. cerevisiae were filtered through a 0.2 µm Acrosdisc syringe filter, then 2 mL of filtrate were removed and replaced with fresh YPD. 100 µL of the counterpart species’ overnight culture was then used to inoculate the tube. After an 18-hour incubation biofilm formation was evaluated through visual inspection.

To test the effects of mannose inhibition on biofilm formation, 15 mL tubes containing 2 mL of YPD and 2 mL of deflocculation buffer (90 mM mannose, 20 mM citrate, 5mM EDTA) were inoculated with 25 µL of DH5α and PMY127 overnight cultures. After an approximately 18-hour incubation, the tubes were inspected for biofilms. Those that were found were scraped onto a microscope slide with a wooden applicator stick for microscopic inspection.

25

Two different methods were used for imaging biofilms containing fluorescently tagged strains. In one, an applicator stick was used to scrape a portion of biofilm off the inside of the tube and onto a microscope slide. In the other, all media was pipetted from the tube and scissors were used cut to the tube lengthwise into slivers approximately .25 cm wide. The portions of biofilm on these slivers were then imaged directly using a Zeis

Imager D1 fluorescent microscope. Time course photos were taken using the second method at two hour intervals.

2.2.3 Long term co-culturing

In the long term co-culturing experiment, cultures of CMY98 S. cerevisiae, derived from the common lab strain S288C, and DH5α E. coli were grown separately from frozen stock overnight in 15 mL plastic culture tubes containing 5 mL of YPD. An initial co- culture tube was made by putting 50 µL of both cultures into a single tube containing 5 mL of YPD. At the same time, monoculture controls were made by putting 50 µL of each strain into its own tube containing 5 mL of YPD. After a three-day incubation, a wooden applicator stick was used to scrape off a small portion of the biofilm which was then used to inoculate a new tube of YPD, establishing the biofilm propagation line. At the same time, 50 µL of media from the same co-culture were taken and used to inoculate a different tube of YPD, establishing the suspension propagation line. 50 µL of media were also taken from the control monocultures and used to inoculate fresh tubes of YPD.

This system of inoculation was repeated every three days for eight cycles for a total of 24

26

days in culture.

At each transfer to fresh media, colonies of each species from the co-culture suspension and biofilm treatments were isolated by dilute plating onto YPD. After incubating overnight, five colonies of each species were isolated, transferred into 5 mL

YPD, and cultured overnight. Then 900 µL of each culture was combined with 900 µL of

30% glycerol solution and frozen at -80° C. Frozen stocks of the control treatments were plated, cultured, and frozen in the same way.

Growth inhibition in spent media was used to infer the amount of overlap in resource usage between species. To measure growth inhibition, all isolated co-culture and control strains were grown from frozen stock in 5 mL YPD overnight. After approximately 20 hours of growth, these cultures were filtered with a 0.2 µm Acrodisc syringe filter. 200 µL of each filtrate was then put into the wells of a flat bottom 96 well plate and inoculated with 25 µL of overnight YPD culture of that filtrate’s counterpart strain. For example, DH5α from the biofilm-selected treatment would be grown in the filtrate of CMY98 from the biofilm-selected treatment. Strains from monoculture controls were grown in the filtrate of the counterpart species’ monoculture control as well.

Additionally, each strain was grown in YPD to establish a baseline growth curve. Each combination of strain and media was replicated in six wells. 96 well plate cultures grew at 30° C for 24 hours in a Tecan Sunrise microplate absorbance reader which agitated the cultures and took optical density measurements every 15 minutes. Using a script written

27

by Colin Maxwell, the maximum growth rate of each well was calculated from the absorbance reader measurements. Growth inhibition was calculated by comparing the maximum growth rates of strains in filtrate to their maximum growth rate in YPD.

2.2.4 TrackScar

Individual S. cerevisiae cell fecundity was measured using the TrackScar method

(Maxwell and Magwene, in press). Populations of live cells were stained for 15 minutes with chitin-binding wheat germ agglutinin, grown at 30° C on a plate shaker for 6 hours, and fixed with an 8% formaldehyde solution. The fixed cells were then stained for 15 minutes with another color of wheat germ agglutinin and imaged with an Applied

Precision DeltaVision fluorescent microscope. The number of daughter cells produced by each imaged cell in a 6-hour period was ascertained by counting the number of bud scars colored by only the second stain.

2.3 Results

2.3.1 DH5α and MACH1 induce flocculation via coaggregation with S. cerevisiae

Flocs were significantly (p<.0001) larger in co-culture than in monoculture. For 11 out of 12 S. cerevisiae strains tested, average floc size was greater when cultured with

DH5α than when cultured alone (Figure 6). The exception to this trend, PMY1506, was very flocculent in monoculture, with an average floc size greater than twice that of the next most flocculent strain. A panel of FLO knockout strains also had significantly

28

(p<.0001) larger flocs when co-cultured with DH5α (Figure 7). No significant difference

(p=.33) in floc size was found when PMY127 was cultured in the filtrate of DH5α, suggesting that secreted substances from E. coli are not responsible for inducing flocculation (Figure 8). Informal observations of the E. coli strain MACH1 in co-culture suggest that it displays flocculation inducing behavior as well. Images of flocs formed by fluorescent strains of MACH1 and PMY1691 show that E. coli cells are present within the flocs (Figure 9).

2.3.2 DH5α and MACH1 form macroscopic biofilms when co-cultured with yeast

When DH5α is co-cultured overnight with S. cerevisiae or C. albicans in a rolling incubator, a robust macroscopic biofilm forms near the liquid-air interface (Figure 10a).

No such biofilm forms in this timeframe when any of these species are cultured in isolation. After three to four days of growth DH5α sometimes forms a weakly visible biofilm in monoculture. The E. coli strain MACH1 also exhibits these behaviors (Figure

10b). Other E. coli strains tested (HB101, OP50, and 10G) do not form macroscopic biofilms (Figure 11a). While all yeast strains tested do form biofilms in co-culture with

DH5α and MACH1, there is strain specific variation in biofilm thickness (Figures 11b and 11c). Biofilm formation, like flocculation, also occurs with various FLO knockout S. cerevisiae strains (Figure 11).

29

30

0F1F2F Figure 6: Average floc size in a panel of S. cerevisiae strains when cultured alone and in the presence of the DH5α strain of E. coli.

31

Figure 7: Average floc sizes of a panel of strains with different FLO genes knocked out. All strains are from the Σ1278B background except the FLO10 knockout, which is from the S288C background.

Figure 8: Average floc size of PMY127 when grown in DH5α overnight culture filtrate and in plain YPD.

Figure 9: An image of a large floc formed by fluorescent strains of MACH1 E. coli (red) and PMY1691 S. cerevisiae (green). Note the E. coli cells interspersed throughout the floc.

3F

32

Figure 10: a. Photo of a co-culture dependent biofilm. On the left is a culture of only DH5α E. coli, on the right is a culture of only PMY127 S. cerevisiae, and in the middle is a co-culture of both with a characteristic biofilm. b. The same as in figure 10a, but with MACH1 instead of DH5α

Figure 11:. a. A co-culture of HB101 E. coli and a strain of C. albicans. Note the lack of biofilm. b. A co-culture of DH5α and the same strain of C. albicans from figure 11a. c. A co-culture of DH5α and a C. albicans strain with low density biofilm forming behavior. d. Biofilms formed by co-culture of MACH1 and FLO8 knockout strains. On the left is a MACH1 monoculture, in the middle is a Σ1278B background ΔFLO8 monoculture, and on the right is an S288C background ΔFLO8 monoculture. In second and fourth positons are co-cultures of MACH1 and Σ1278B ΔFLO8 and S228C ΔFLO8, respectively.

33

2.3.3 Biofilms are an intimate association of both species

Images of mature 20 hour old biofilms scraped onto a microscope slide show that both species are present in the biofilm (Figure 12a). In situ imaging shows the same, albeit with more heterogeneity in their spatial distribution (Figure 12b). Time courses show that the biofilm contains both species at the earliest stages of formation (Figure 13).

These results indicate that coaggregation and attachment to the tube wall are concurrent processes.

34

Figure 12:. a. Images of a biofilms formed by a fluorescently tagged strain of MACH1 (red) and PMY1690 (green) that has been scraped from the inside of the tube onto a microscope slide. b. In situ images of biofilms formed by the same strains. Note the presence of areas where one species predominates.

35

36

Figure 13: A time course of biofilm formation in co-cultures of fluorescently tagged MACH1 (red) and PMY1690 (green). Note the presence of both E. coli and S. cerevisiae cells in the first time point. No adhering cells were detectable at 2 hours post inoculation (not shown).

2.3.4 Biofilm formation requires live E. coli and is affected by deflocculation buffer

No biofilms formed when S. cerevisiae was grown in the filtrate of E. coli or vice versa. Similarly, no biofilms formed when S. cerevisiae was grown in the presence of heat killed E. coli. Diffuse biofilms did form when MACH1 or DH5α E. coli was grown in the presence of heat killed S. cerevisiae (data not shown). These results indicate that the cellular material of both species must be present for the formation of a macroscopic biofilm, but only E. coli must be alive. Biofilms formed in a 1:1 mixture of YPD and deflocculation buffer are much thinner than normal, and microscopy shows that they contain very few S. cerevisiae cells (Figure 14). This suggests that yeasts’ adhesion to the biofilm is mannose-mediated.

Figure 14: Comparison of a biofilm formed in plain YPD (left) and a biofilm formed in a mixture of YPD and deflocculation buffer (right). Note the much lower density of yeast cells in the deflocculation treatment.

37

2.3.5 Selection for biofilm association in S. cerevisiae leads to slower growth rates

No consistent trend in cross species growth inhibition was seen in strains that were continuously co-cultured (data not shown). However, the biofilm-selected CMY98

S. cerevisiae co-culture strain did develop a lower maximum growth rate in YPD than either the monocultured or suspension-propagated co-culture strains (Figure 15).

TrackScar measurements confirmed that after seven selection events for biofilm association CMY98 cells were on average significantly (p < .0001) less fecund than pre- selection cells (Figure 16a). This was due to a shift in the mode of reproduction rates rather than an increase in the number of non-reproducing cells (Figure 16b).

Figure 15: Maximum growth rates in YPD of CMY98 strains from co-culture selection experiment

38

Figure 16:. a. Comparison of the average individual fecundity in YPD of the CMY98 starting population and the population after seven selection events for biofilm association. b. Distributions of individual fecundity for of those same populations. 39

2.4 Discussion

A variety of bacterial species, including E. coli, have shown the ability to coaggregate with S. cerevisiae and C. albicans cells (Mirelman, Altmann, and Eshdat

1980). This phenomenon has even been exploited in a treatment for diarrhea called anti- adhesion therapy, where S. cerevisiae cells are used to competitively inhibit bacterial cells’ adhesion to mucus membranes (Peng et al. 2001). The increased floc size found in co-cultures of S. cerevisiae and MACH1 or DH5α appear to be another example of this coaggregation behavior. The putative agent of this coaggregation is the mannose- binding FimH protein present on E. coli fimbriae (Krogfelt, Bergmans, and Klemm 1990;

Peng et al. 2001). A cell wall-anchored E. coli organelle causing coaggregation fits with the data shown here: it would not be present in culture filtrates (Figure 8), would cause

E. coli cells to be enmeshed within flocs (Figure 9), and would obviate the role of S. cerevisiae adhesins in creating flocs (Figure 7).

Plastic adhering biofilms of E. coli and S. cerevisiae, whether through oversight or rarity, appear to be absent from the literature. However, a very similar phenomenon involving Lactobacillus plantarum and S. cerevisiae has been described (Furukawa et al.

2011). In their study, Furukawa et al. concluded that a bacterial mannose-binding protein was responsible for cell-cell adhesion, but that a separate mechanism was responsible for cell-plastic adhesion. I propose that the same is true for the E. coli and S.

40

cerevisiae biofilms. Biofilm formation is a binary and strain specific trait in E. coli, but all yeast strains can form them, albeit with variations in thickness. Biofilms can still form with heat-killed S. cerevisiae, but not with heat-killed E. coli – as would be expected if formation depended upon an easily denatured E. coli protein adhering to a robust S. cerevisiae polysaccharide. When free mannose is added to normally robust biofilm- inducing conditions, thin biofilms containing few yeast cells form. All these results indicate that mannose dependent, E. coli driven adhesion to S. cerevisiae is the main determinant in creating a robust biofilm. The formation of an E. coli dominated biofilm when mannose is added to the media also indicates that biofilm plastic adhesion is an E. coli-mediated trait and that it is not reliant on the same mechanism as cell-cell adhesion.

Going ahead, these hypotheses should be tested by knocking out fimH in MACH1 and

DH5α to see if it abolishes biofilm formation, as well as replacing fimH alleles in non- biofilm forming E. coli strains with those of biofilm-forming ones to see if it induces the behavior. Supplementing media with sucrose palmitate, which is known to inhibit bacterial biofilm formation, and measuring the effects on floc size and biofilm formation would test if coaggregation and plastic adhesion happen through separate mechanisms

(Furukawa et al. 2011). It would also be worthwhile to investigate the basis of biofilm thickness variation among yeast strains. One explanation could be that faster reproducing strains create more biomass once attached to the nascent biofilm. Another is

41

that strains with more mannose in their cell walls have higher binding affinities which result in a greater presence in the biofilm.

The most obvious adaptive function of forming a biofilm at the liquid-air interface is to ensure access to oxygen and thus metabolically efficient respiration. If E. coli are the initial actors in creating these biofilms, it is puzzling why they would include yeast, which would compete with them for resources and potentially create hypoxic conditions (Fox et al. 2014). It could be that the addition of yeasts’ biomass makes the biofilm sturdier, reducing the chances of E. coli cells being swept away from their favorable location. Similarly, an enclosure of yeast cells could provide a favorable microenvironment. Such an arrangement has been seen in co-cultures of C. albicans and the anaerobic Clostridium perfingens. In ambient oxic conditions C. perfingens induces aggregation in C. albicans and then grows in the hypoxic microenvironment inside those aggregates (Fox et al. 2014). E. coli could benefit from yeast aggregation in a similar manner – an E. coli cell surrounded by yeast cells is presumably less likely to be exposed to bacterivores or toxins than one out in the open, or even one surrounded by other E. coli. Conversely, coaggregation with yeast cells could provide no fitness advantage and simply be a result of E. coli having non-specific mannose binding adhesins. Future experiments comparing the growth of E. coli in yeast co-cultures where it is allowed to coaggregate and where it is prevented with chemical agents could shed light on the existence of a fitness advantage conferred by the behavior.

42

Observing the phenotypic changes that occur when species evolve in proximity to one another is one way to infer the nature of their interactions. The lower growth rate seen in S. cerevisiae populations selected for biofilm adhesion suggests that there are metabolic costs involved in living in the biofilm. A crowded environment could lead to exposure to high levels of toxic metabolic wastes, low oxygen levels, or unfavorable pH conditions. S. cerevisiae exposed to these conditions would have to deal with them using metabolic resources that would otherwise go towards reproduction, ultimately resulting in a hardy but slow growing population. This resource usage tradeoff could occur even if conditions in the biofilm are more favorable than those in suspension. If residence in the biofilm is contingent on E. coli binding to S. cerevisiae cell wall polysaccharides, S. cerevisiae cells that produce more of them would be selected for. A metabolic tradeoff between surface polysaccharide production and growth would also lead to the lower growth rate seen in biofilm selected populations. One explanation that can be discarded is the evolution of facilitation – specifically that biofilm S. cerevisiae became reliant upon

E. coli metabolic products and then grew slowly in their absence. If this was the case, higher growth rates in E. coli filtrate over the course of selection would have been seen, but they were not (data not shown). Bulk segregant analyses between biofilm selected, suspension selected, and monocultured populations could shed light on the genes involved in these adaptations. Another question that remains is whether slower growth is an adaptation specific to living in biofilms with E. coli or just biofilms in general. This

43

could be addressed by selecting populations for biofilm formation in monoculture and co-culture and comparing their growth rates. Since S. cerevisiae does not form interface biofilms under typical monoculture conditions, the selection could instead be for large flocs, which have similar microenvironments to biofilms and can be found in both monoculture and co-culture (Fox et al. 2014).

44

3. Filamentation in E. coli

3.1 Introduction

The study of filamentation in bacteria can be traced back to the early decades of the

20th century, when debate raged among bacteriologists over the morphology and life cycles of bacteria. In one camp were the monomorphists, who asserted that each bacterial species exhibited a single morphology with only minor variation among individuals. On the other side were the pleomorphists, who held that not only could bacteria of the same species take on a variety of morphologies, but also that bacteria had complex lifecycles that included single cell forms, plasmodial aggregates, and

(Almquist 1922). Some even went as far as to suggest that bacteria were simply a stage of the fungal lifecycle (Bergstrand 1920). While the more extreme views of the pleomorphist school have long been discarded by the scientific community, a multitude of studies since those early days have shown that bacteria are capable of taking on variant morphologies (Wainright 1997). The most well studied of these morphologies is filamentation.

Filamentation has been observed in a number of bacterial species in response to various stressors. Factors shown to induce filamentation include exposure to UV radiation, oxidative radicals, urea, high pressure, acidity, low temperatures, , starvation, and host immune responses (Janion et al. 2002; Wilson 1905; Miller et al.

2004; Wainwright et al. 1999; Justice et al. 2006; Jones, Vail, and McMullen 2013).

45

Filamentation has been shown to make cells resistant to both macrophage phagocytosis and antibiotics, making the phenomenon especially relevant to the medical field. In E. coli filamentation is associated with the SOS response, which is a suite of physiological reactions triggered by DNA damage. When a cell experiences DNA damage, transcription of over 30 unlinked genes begins. This suite of genes acts to both repair

DNA damage and inhibit fission. Among the genes transcribed is sulA, which codes for a protein that binds to and inhibits the action of the division protein FtsZ. In the absence of FtsZ, a cell grows but cannot divide, leading to the formation of a non-septate filament. Once DNA repair is complete, the SOS gene suite is repressed and the filament begins to divide (Justice et al. 2008). Although the SOS response is canonically induced by DNA stress, filamentation has been shown to be induced by other factors such as antibiotics that disrupt peptidoglycan synthesis and the excreted products of protist bacterivores (Corno and Jürgens 2006; Miller et al. 2004).

While studying the evolutionary dynamics of mixed species biofilms, I observed that E. coli cells grown in the presence of their own culture filtrates were often markedly

(>10X) longer than their counterparts grown in fresh media. Follow up experiments ruled out pH stress and nutrient depletion as the main inducers of filamentation, suggesting that E. coli excrete a substance that induces their own filamentation. Here I present findings on the production and action of this putative filamentation-inducing agent (PFIA).

46

3.2 Methods

3.2.1 Culturing

Bacterial cultures were grown in plastic 15 mL culture tubes containing 5 mL of either 2% glucose YPD, 2% glucose LB, or LB overnight in a rolling incubator at 30° C.

After approximately 20 hours of growth, cultures were filtered with a 0.2 µm Acrodisc syringe filter. For experimental treatments, 2.5 mL of a given filtrate were combined with 2.5 mL of fresh media and inoculated with 25 µL of overnight bacterial culture. For control treatments, either 5 mL of fresh media or 2.5 mL of sterile water mixed with 2.5 mL of fresh media were inoculated with 25 µL of overnight bacterial culture. Before inoculation the pH of every treatment was measured with a Denver Instrument pH meter and adjusted with HCl or NaOH as necessary for the experiment. After incoculation cultures were incubated at 30° C overnight and imaged.

3.2.2 Imaging and analysis

Cells were imaged using an AMG Evos microscope at 40X magnification. For each biological replicate, seven images were taken from a 2 µL sample drawn from the middle of the culture tube. Between 3 and 20 biological replicates were imaged for each treatment. Samples with low cellular densities were centrifuged for 5 minutes at 3220 G and resuspended in 200 µL of media before imaging. After taking an image the stage was moved randomly with the screen off to prevent bias. Images were processed using

Fiji’s “Analyze Particles” function with size range set from 0-infinity and circularity set

47

from 0-1 (Schindelin et al. 2012). Tests for significance were done using one-way

ANOVAs in R and figures were made using the ggplot2 package (Wickham, 2009).

3.3 Results

3.3.1 E. coli can produce a signal that induces their own filamentation

When E. coli cells are exposed to certain stressors, they can develop a filamentous phenotype with a length 5-60 times longer than normal (Figure 17). Filamentous cells can be found at low densities under normal culturing conditions, but become more common at low pH and nutrient levels. When MACH1 is cultured in a 1:1 mixture of its own overnight YPD filtrate and fresh YPD, filamentous cells become abundant and the average size of the largest decile of cells (ASLD) more than quadruples (Figure 18).

ASLD is used here instead of whole population averages because it is a more sensitive method for detecting changes in filamentation and excludes erroneous measurements of small debris (Jones, Vail, and McMullen 2013). YPD filtrates are mildly acidic (pH 5.1-

5.3) and nutrient depleted. When these stressors are introduced in the absence of filtrate, ASLD rises significantly (p<.0001), but only 13% as much as when filtrate is present (Figure 18). S. cerevisiae filtrate also raises ASLD, but only 24% as much as

MACH1 filtrate, indicating that other factors introduced by general microbial metabolism are not primarily responsible for the increase. These data indicate that a metabolic product of MACH1 induces filamentation in its own cells.

48

Figure 17: A comparison of normal (top) and filamentous (bottom) MACH1 cells at 40X magnification.

49

Figure 18: ASLDs of MACH1 after overnight culture in a variety of media conditions

3.3.2 There is variation in filamentation behavior among strains

There is variation in the magnitude of filtrate-induced filamentation between the five strains of E. coli tested. DH5α has a remarkably strong filamentation response with an approximately 1100% increase in ASLD when cultured in its own filtrate (Figure 19).

MACH1 is the second strongest responder with an approximately 350% increase. HB101,

10G, and OP50 all show less than a 70% increase. These differences indicate that the mechanisms governing filamentation are polymorphic among E. coli strains.

50

Figure 19: ASLDs of all five members of the E. coli strain panel when cultured in YPD and their own filtrates

3.3.3 Strain variation in filamentation is due to differences in reaction to inducing agent, not production of it

When the high-responding MACH1 strain was cultured in the filtrate of the low- responding HB101 strain its ASLD was slightly (1.7 µm2) higher than when it was cultured in its own filtrate (Figure 20). When HB101 was cultured in MACH1’s filtrate, its ASLD was slightly lower (0.8 µm2) than when it was cultured in its own. In its own filtrate, HB101’s ASLD only exceeded the nutrient and acidity stressed control by 0.2

51

µm2, and in MACH1’s filtrate it was 0.6 µm2 lower than the stressed control. These results suggest that both strains produce the PFIA, but differ in their response to it.

Figure 20: MACH1 and HB101 ASLDs when cultured in one another’s filtrates, their own filtrates, plain YPD, and nutrient and acidity stressed controls.

3.3.4 The PFIA is heat stable up to 80°C

MACH1 filtrate remains capable of inducing filamentation after being heated to

80° C for one hour (Figure 21). The ASLD of MACH1 cells grown in heat-treated filtrate was within 1.6 µm2 of cells grown in non-heated filtrate, while the nutrient and acidity stressed control was over 14.5 µm2 lower. This indicates that the putative filamentation- inducing agent remains active when exposed to high heat, suggesting that is not proteinaceous.

52

Figure 21: ASLDs of MACH1 cells grown in heat-treated and typical filtrate mixtures and control conditions

3.3.5 Glucose is necessary for the production of the PFIA

MACH1 cells did not show any filamentation behavior when grown in filtrate mixtures made from overnight cultures in LB media (Figure 22). LB does not contain any glucose, and the main carbon source is catabolizable amino acids. Ammonium is released during amino acid catabolysis, resulting in an alkaline filtrate (Sezonov et al.,

2007). When LB is supplemented with 2% glucose, overnight cultures become acidic, with filtrate mixtures at approximately pH 5.2 (data not shown). 2% glucose LB filtrate is capable of inducing filamentation at levels similar to YPD filtrate, with less than a 1.5

µm2 difference between their ASLDs. These results suggest that glucose metabolism is necessary for the creation of the PFIA.

53

Figure 22: ASLDs of MACH1 cells in LB, LB 2% glucose, and YPD media and filtrates

3.3.6 Low pH conditions and acclimatization are necessary for filamentation to occur

Raising the pH of YPD filtrate from 5.2 to 7.0 at the time of inoculation abolished its filamentation-inducing effect in MACH1 (Figure 23). This suggests that cells are not sensitive to the PFIA at neutral pH and that the conditions at inoculation are a major determinant of cell morphology 20 hours later. When YPD filtrates were inoculated with

MACH1 cells cultured overnight in LB, filamentation was greatly reduced, with an

ASLD approximately 13 µm2 less than in filtrates inoculated with overnight YPD cultures (Figure 24). This result suggests that cells from YPD cultures, which are acidic and contain PFIA, are “primed” for filamentation in a way that cells from LB cultures are not.

54

Figure 23: ASLDs of MACH1 control and filtrate treatments with variable pH at time of inoculation

Figure 24: ASLDs of MACH1 controls and filtrate treatments with inoculums from LB or YPD cultures

55

3.4 Discussion

Historically filamentation was associated with sick and dying cells, but is now considered a widely adaptive behavior (Justice et al. 2008). Filamentous cells have been shown to be resistant to phagocytosis, heat, alkalinity, and antibiotics (Miller et al. 2004;

Wortinger, Quardokus, and Brun 1998; Justice et al. 2004). The filamentation behavior displayed here is likely also an adaptive reaction to adverse conditions. The interesting aspect of this case is that intercellular signaling appears to play a larger role in inducing filamentation than the adverse conditions themselves.

The obvious stressor introduced by growth in spent media is nutrient limitation, which has been shown to induce filamentation in E. coli (Wainwright et al. 1999). Less obvious and particular to media containing carbohydrates is low pH, which has also been shown to induce filamentation, albeit at lower levels (4.1-4.4) than seen here (5.1-

5.3) (Mattick, Rowbury, and Humphrey 2003). While these stressors together cause

MACH1 cell size to increase significantly (p<.0001), they only account for 13% of the increase in ASLD seen when cells are cultured in filtrate. The majority of filamentation induction must then be due to another factor introduced by filtrate other than decreased nutrient and pH levels. This factor, the PFIA, has several of the characteristic qualities of a quorum sensing molecule described in Williams et al. (2007). It is only produced in the presence of glucose (Figure 22), and thus only under certain physiological or environmental circumstances. It is contained in filtered media, and thus accumulates in

56

the extracellular environment. Most importantly, filamentation induction is contingent on factors other than the presence of the PFIA itself (Figures 23 and 24), satisfying the requirement that the cellular response go beyond the physiological activities necessary to metabolize the molecule. I propose that the PFIA is a quorum sensing molecule that causes filamentation to occur under acidic and/or nutrient limited conditions.

Bacterial quorum sensing is mediated by molecules called autoinducers, of which there are three main types: acyl-homoserine lactone (AHL), oligopeptide, and furanosyl borate diester (Bassler 2002). AHLs are commonly employed by Gram-negative bacteria, and E. coli can detect them using the protein SdiA, but no AHL-producing strains of E. coli have been identified (Williams et al. 2007). Oligopeptide autoinducers are employed by Gram-positive bacteria, and would not be expected to be produced by E. coli

(Schauder, Stephan and Bassler 2001). There is only one known furanosyl borate diester autoinducer, autoinducer-2 (AI-2). AI-2 in reality is a collection of related furanone derivatives that spontaneously cyclize from the precursor 4,5-dihydroxy- 2,3- pentanedione (DPD) (Williams et al. 2007). Production of DPD and sensitivity to AI-2 is found in a wide variety of both Gram-positive and Gram-negative bacteria, including E. coli, and has been speculated to be the basis of widespread species-nonspecific bacterial communication (Delisa, Valdes, and Bentley 2001; Bassler 2002).

AI-2 has several qualities aside from being produced by E. coli that make it a likely candidate for the PFIA. It has been found to be heat-stable to 80° C, which would

57

account for heat-treated filtrates’ ability to induce filamentation (Figure 21) (Surette and

Bassler 1998). Glucose has been found to increase extra-cellular AI-2 fivefold (Delisa,

Valdes, and Bentley 2001). The physiological basis for this increase is catabolite repression of the lsr operon that codes for the transporter necessary for internalizing AI-

2. The lsr operon is under control of the cyclic adenosine monophosphate (cAMP)- catabolite activator protein (CAP) system, and in the presence of glucose cAMP cannot activate CAP, which in turn fails to activate transcription of the lsr operon. Cells are then unable to take in AI-2, and it accumulates in the extra-cellular environment (Xavier and

Bassler 2005). This could account for the difference in filamentation induction between filtrates made with cultures containing glucose and those without (Figure 22).

Additionally, a survey of 44 E. coli strains showed that only 38% had complete lsr operons, while all but one had a functional luxS gene (Brito et al. 2013). This fits with the observation that the filtrate of HB101 can induce strong filamentation in MACH1, but not vice versa (Figure 20). If the PFIA were AI-2, it would be expected to be secreted by all strains, but only have an effect on a minority with functioning lsr operons.

Sequencing the lsr operon of all strains could reveal if it is the determining factor in filamentation behavior. Most significantly, AI-2 upregulates several genes involved in the SOS response and has been speculated to directly induce it (Delisa, Valdes, and

Bentley 2001; Sperandio et al. 2001). It has also been found to stimulate biofilm formation in E. coli (Zuo et al. 2006). The two highly filamenting strains DH5α and

58

MACH1 are also the only strains of the E. coli panel to form robust biofilms with yeast

(Figures 10a and 10b), and both behaviors could potentially be explained by sensitivity to AI-2. These examples are by no means conclusive, and there are counterexamples to suggest that AI-2 is not the PFIA, the strongest of which is a report that DH5α, the strongest filamenting strain, does not produce AI-2 when cultured in 0.5% glucose LB

(Surette and Bassler 1998). Luckily, methods for quantifying and manipulating AI-2 levels are well established, and going forward these methods can be used to test if AI-2 is the PFIA.

Regardless of the identity of the PFIA, some tentative conclusions about its production and activity can be drawn from my experiments. Glucose and/or the acidic pH that develops from its metabolization is necessary for PFIA production or accumulation (Figure 22). LB cultures of MACH1 have OD600 values approximately twice that of cultures containing glucose (data not shown), ruling out the hypothesis that more PFIA is created in glucose-containing media due to greater cell density.

Filamentation is only induced in filtrate cultures that are acidic (pH 5.3) at the time of inoculation (Figure 23) and is only seen when filtrate cultures are inoculated with cells that were cultured in PFIA-inducing conditions (Figure 24). This suggests that both an acidic, PFIA-containing environment and previous acclimatization to a similar environment are necessary for filamentation to occur. High levels of PFIA could signal to cells a crowded, possibly nutrient deficient environment. Once inoculated into filtrate,

59

a relatively small number of cells would be exposed to a PFIA concentration characteristic of a much larger population, increasing the amount of PFIA molecules individual cells are exposed to. Coupled with pH stress, this would be interpreted as a very crowded and inhospitable environment, leading to activation of the SOS response and filamentation. This sequence of events is highly speculative, and future experiments should be performed to test it. A good starting point would be comparing pH, glucose levels, and filamentation induction over time to determine what the environmental thresholds for PFIA production are.

Filamentous morphology and quorum sensing have both been directly linked to bacterial pathogenicity and immune system subversion (Klein et al. 2015; Rutherford and Bassler 2012). In light of their medical relevance, thorough investigation of these phenomena is warranted. The molecular arsenal that we have used to control bacteria for a century is becoming obsolete, and new methods of control must be developed. If we can understand the mechanisms that govern bacterial behavior at the level of both the individual and population, we may be able to avoid the worst of the coming “post- antibiotic” era.

60

4. Conclusions

My research provides bases for multiple lines of future investigation. Extending the results of the plastic adhesion QTL experiment is a matter of sequencing existing

DNA libraries and applying an established analysis pipeline. Cross verification between populations will produce a list of strongly supported candidate genes for which the cost of experimental analysis can be justified. The plastic adhesion assay used here can be employed to measure the effects of overexpression and deletion of implicated genes, although more sophisticated measures of plastic adhesion exist and should also be considered (Kang and Choi 2005). There are multiple directions to take the mixed species biofilm project, but dissecting the genetic basis of biofilm-selected S. cerevisiae’s slow growth is the one for which the Magwene lab is most well-equipped. The results from such research could potentially shed light on the metabolic tradeoffs involved not only in mixed S. cerevisiae and E. coli biofilms but fungal biofilms in general.

Manipulations of the fimH gene in E. coli should also be performed as they could reveal a physiological basis for the strain-specific formation of macroscopic mixed species biofilms. The most promising direction for the E. coli filamentation project is an investigation into the action of AI-2 in filamentation. An established assay utilizing

Vibrio fischeri’s AI-2-dependent fluorescence can be used to quantify differences in AI-2 production by E. coli under PFIA and non-PFIA inducing conditions (Taga and Xavier

2011). Purified DPD, AI-2’s precursor molecule, can also be added to media so that AI-

61

2’s effects on filamentation can be observed directly and in isolation from other factors present in culture filtrate. DPD should also be added to E. coli and S. cerevisiae co- cultures to see if it affects the thickness of mixed species biofilms. Additionally, sequencing the lsr and luxS genes of the E. coli panel strains could establish a mechanistic basis for the variation in filamentation seen between strains. With these additional experiments, the projects I have started can provide considerable insights into a variety of microbial behaviors and ultimately improve our abilities to manipulate them.

62

References

Adam, Berit, George S. Baillie, and L. Julia Douglas. 2002. “Mixed Species Biofilms of Candida Albicans and Staphylococcus Epidermidis.” Journal of Medical 51: 344–49.

Almquist, Ernst. 1922. “Variation and Life Cycles of .” Médecine et Maladies Infectieuses 1 (6): 109–12. doi:10.1016/S0399-077X(71)80055-0.

Archer, Colin T., Jihyun F. Kim, Haeyoung. Jeong, Jin Park, Claudia E. Vickers, Sang Lee, and Lars K. Nielsen. 2011. “The Genome Sequence of E. Coli W (ATCC 9637): Comparative Genome Analysis and an Improved Genome-Scale Reconstruction of E. Coli.” BMC Genomics 12 (1): 9. doi:10.1186/1471-2164-12-9.

Bachman, Barbara J. 1972. “Pedigrees of Some Mutant Strains of Escherichia Coli K12.” Bacteriological Reviews 36 (4): 525–57. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC408331/pdf/bactrev00201- 0120.pdf\npapers2://publication/uuid/58180B42-811B-442F-92F4-DF6306559A6B.

Bassler, Bonnie L. 2002. “Small Talk: Cell-to-Cell Communication in Bacteria.” Cell 109 (4): 421–24. doi:10.1016/S0092-8674(02)00749-3.

Bayly, Jennifer C., Lois M. Douglas, Isak S. Pretorius, Florian F. Bauer, and Anne M. Dranginis. 2005. “Characteristics of Flo11-Dependent Flocculation in Saccharomyces Cerevisiae.” FEMS Yeast Research 5 (12): 1151–56. doi:10.1016/j.femsyr.2005.05.004.

Benachour, Abdellah, György Sipos, Isabelle Flury, Fulvio Reggiori, Elisabeth Canivenc- Gansel, Christine Vionnet, Andreas Conzelmann, and Mohammed Benghezal. 1999. “Deletion of GPI7, a Yeast Gene Required for Addition of a Side Chain to the Glycosylphosphatidylinositol (GPI) Core Structure, Affects GPI Protein Transport, Remodeling, and Cell Wall Integrity.” Journal of Biological Chemistry 274 (21): 15251– 61. doi:10.1074/jbc.274.21.15251.

Bergstrand, Hilding. 1920. “On the Nature of Bacteria.” Médecine et Maladies Infectieuses 1 (6): 109–12. doi:10.1016/S0399-077X(71)80055-0.

Bony, Muriel, Denise Thines-sempoux, Pierre Barre, and Buno Blondin. 1997. “Localization and Cell Surface Anchoring of the the Saccharomyces Cerevisiae Flocculation Protein Flo1p.” Microbiology 179 (15): 4929–36.

63

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=179343&tool=pmcentre z&rendertype=abstract.

Brito, Patrícia H., Eduardo P C Rocha, Karina B. Xavier, and Isabel Gordo. 2013. “Natural Genome Diversity of AI-2 Quorum Sensing in Escherichia Coli: Conserved Signal Production but Labile Signal Reception.” Genome Biology and Evolution 5 (1): 16–30. doi:10.1093/gbe/evs122.

Cherry, J. Michael, Eurie L. Hong, Craig Amundsen, Rama Balakrishnan, Gail Binkley, Esther T. Chan, Karen R. Christie, et al. 2012. “Saccharomyces Genome Database: The Genomics Resource of Budding Yeast.” Nucleic Acids Research 40 (D1): 700–705. doi:10.1093/nar/gkr1029.

Choi, Yongwook. 2012. “A Fast Computation of Pairwise Sequence Alignment Scores between a Protein and a Set of Single-Locus Variants of Another Protein.” Proceedings of the ACM Conference on Bioinformatics, Computational Biology and Biomedicine - BCB ’12, 414–17. doi:10.1145/2382936.2382989.

Corno, Gianluca, and Klaus Jürgens. 2006. “Direct and Indirect Effects of Protist Predation on Population Size Structure of a Bacterial Strain with High Phenotypic Plasticity.” Applied and Environmental Microbiology 72 (1): 78–86. doi:10.1128/AEM.72.1.78.

Cullen, Paul J., and George F. Sprague. 2000. “Glucose Depletion Causes Haploid Invasive Growth in Yeast.” Proceedings of the National Academy of Sciences of the United States of America 97 (25): 13619–24. doi:10.1073/pnas.240345197.

Delisa, Matthew P., James J. Valdes, and William E. Bentley. 2001. “Mapping Stress- Induced Changes in Autoinducer AI-2 Production in Chemostat-Cultivated Escherichia Coli K-12 Mapping Stress-Induced Changes in Autoinducer AI-2 Production in Chemostat-Cultivated Escherichia Coli K-12.” Journal of Bacteriology 183 (9): 2918–28. doi:10.1128/JB.183.9.2918.

Dunne, W. Michael. 2002. “Bacterial Adhesion: Seen Any Good Bioflms Lately?” Clinical Microbiology Reviews 15 (2): 155–66. doi:10.1128/CMR.15.2.155.

Fidalgo, Manuel, Ramon Barrales, and Juan Jimenez. 2009. “Characterization of Chromosomal Integration Sites for Heterologous Gene Expression in Saccharomyces Cerevisiae.” Yeast 26 (10): 545–51. doi:10.1002/yea.

64

Fox, Emily P., Elise S. Cowley, Clarissa J. Nobile, Nairi Hartooni, Dianne K. Newman, and Alexander D. Johnson. 2014. “Report Anaerobic Bacteria Grow within Candida Albicans Biofilms and Induce Biofilm Formation in Suspension Cultures.” Current Biology 24 (20). Elsevier Ltd: 2411–16. doi:10.1016/j.cub.2014.08.057.

Furukawa, Soichi, Natsumi Nojima, Kanako Yoshida, Satoru Hirayama, Hirokazu Ogihara, and Yasushi Morinaga. 2011. “The Importance of Inter-Species Cell-Cell Co-Aggregation between Lactobacillus Plantarum ML11-11 and Saccharomyces Cerevisiae BY4741 in Mixed-Species Biofilm Formation.” Bioscience, Biotechnology, and Biochemistry 75 (8): 1430–34. doi:10.1271/bbb.100817.

Gagiano, Marco, Dewald van Dyk, Florian F. Bauer, Marius G. Lambrechts, and Isak S. Pretorius. 1999. “Msn1p/ Mss10p, Mss11p and Muc1p/ Flo11p Are Par T of a Signal Transduction Pathway Downstream of Mep2p Regulating Invasive Growth and Pseudohyphal Differentiation in Saccharomyces Cerevisiae.” Molecular Microbiology 31: 103–16.

Granek, Joshua A., Debra Murray, Ömür Kayrkçi, and Paul M. Magwene. 2013. “The Genetic Architecture of Biofilm Formation in a Clinical Isolate of Saccharomyces Cerevisiae.” Genetics 193 (2): 587–600. doi:10.1534/genetics.112.142067.

Guo, Bing, Cora A. Styles, Qinghua Feng, and Gerald R. Fink. 2000. “A Saccharomyces Gene Family Involved in Invasive Growth, Cell-Cell Adhesion, and Mating.” Proceedings of the National Academy of Sciences of the United States of America 97 (22): 12158–63. doi:10.1073/pnas.220420397.

Janion, Celina, Anna Sikora, Anetta Nowosielska, and Elzbieta Grzesiuk. 2002. “Induction of the SOS Response in Starved Escherichia Coli.” Environmental and Molecular Mutagenesis 40 (2): 129–33. doi:10.1002/em.10094.

Jarosz, Daniel F., Jessica C. S. Brown, Gordon A. Walker, Manoshi S. Datta, W. Lloyd Ung, Alex K. Lancaster, Assaf Rotem, et al. 2014. “Cross-Kingdom Chemical Communication Drives a Heritable, Mutually Beneficial Prion-Based Transformation of Metabolism.” Cell 158 (5): 1083–93. doi:10.1016/j.cell.2014.07.025.

Jones, Tineke H., Kathleen M. Vail, and Lynn M. McMullen. 2013. “Filament Formation by Foodborne Bacteria under Sublethal Stress.” International Journal of Food Microbiology 165 (2): 97–110. doi:10.1016/j.ijfoodmicro.2013.05.001.

Justice, Sheryl S., Chia Hung, Julie A. Theriot, Daniel A. Fletcher, Gregory G. Anderson,

65

Matthew J. Footer, and Scott J. Hultgren. 2004. “Differentiation and Developmental Pathways of Uropathogenic Escherichia Coli in Urinary Tract Pathogenesis.” Proceedings of the National Academy of Sciences of the United States of America 101 (5): 1333–38. doi:10.1073/pnas.0308125100.

Justice, Sheryl S., David A. Hunstad, Lynette Cegelski, and Scott J. Hultgren. 2008. “Morphological Plasticity as a Bacterial Survival Strategy.” Nature Reviews Microbiology 6 (2): 162–68. doi:10.1038/nrmicro1820.

Justice, Sheryl S., David A. Hunstad, Patrick C. Seed, and Scott J. Hultgren. 2006. “Filamentation by Escherichia Coli Subverts Innate Defenses during Urinary Tract Infection.” Proceedings of the National Academy of Sciences of the United States of America 103 (52): 19884–89. doi:10.1073/pnas.0606329104.

Kang, Seoktae, and Heechul Choi. 2005. “Effect of Surface Hydrophobicity on the Adhesion of S. Cerevisiae onto Modified Surfaces by Poly(styrene-Ran-Sulfonic Acid) Random Copolymers.” Colloids and Surfaces B: Biointerfaces 46 (2): 70–77. doi:10.1016/j.colsurfb.2005.08.017.

Kapteyn, Johan C., Herman Van Den Ende, and Frans M. Klis. 1999. “The Contribution of Cell Wall Proteins to the Organization of the Yeast Cell Wall.” Biochimica et Biophysica Acta - General Subjects 1426 (2): 373–83. doi:10.1016/S0304-4165(98)00137- 8.

Klein, Kasper, Yaseelan Palarasah, Hans J. Kolmos, Jakob Møller-Jensen, and Thomas E. Andersen. 2015. “Quantification of Filamentation by Uropathogenic Escherichia Coli during Experimental Bladder Cell Infection by Using Semi-Automated Image Analysis.” Journal of Microbiological Methods 109: 110–16. doi:10.1016/j.mimet.2014.12.017.

Kobayashi, Osamu, Nobuyuki Hayashi, Ryota Kuroki, and Hidetaka Sone. 1998. “Region of Flo1 Proteins Responsible for Sugar Recognition.” Journal of Bacteriology 180 (24): 6503–10.

Kolter, Roberto, and E. Peter Greenberg. 2006. “Microbial Sciences: The Superficial Life of Microbes.” Nature 441 (7091): 300–302. doi:10.1038/441300a.

Krogfelt, Karen A., Hans Bergmans, and Per Klemm. 1990. “Direct Evidence That the FimH Protein Is the Mannose Specific Adhesin of Escherichia Coli Type 1 Fimbriae.” Infection and Immunity 58 (6): 1995–99.

66

Lawrence, Diane, Francesca Fiegna, Volker Behrends, Jacob G. Bundy, Albert B. Phillimore, Thomas Bell, and Timothy G. Barraclough. 2012. “Species Interactions Alter Evolutionary Responses to a Novel Environment.” PLoS Biology 10 (5). doi:10.1371/journal.pbio.1001330.

Lo, Wan-Sheng, and Anne M. Dranginis. 1998. “The Cell Surface Flocculin Flo11 Is Required for Pseudohyphae Formation and Invasion by Saccharomyces Cerevisiae.” of the Cell 9 (1): 161–71. doi:10.1091/mbc.9.1.161.

Madhani, Hiten D., and Gerald R. Fink. 1997. “Combinatorial Control Required for the Specificity of Yeast MAPK Signaling.” Science 275 (5304): 1314–17. doi:10.1126/science.275.5304.1314.

Magwene, Paul M., John H. Willis, and John K. Kelly. 2011. “The Statistics of Bulk Segregant Analysis Using next Generation Sequencing.” PLoS Computational Biology 7 (11): 1–9. doi:10.1371/journal.pcbi.1002255.

Mattick, Karen L., Robin J. Rowbury, and Tom J. Humphrey. 2003. “Morphological Changes to Escherichia Coli O157:H7, Commensal E. Coli and Salmonella Spp in Response to Marginal Growth Conditions, with Special Reference to Mildly Stressing Temperatures.” Science Progress 86 (Pt 1-2): 103–13. http://www.ncbi.nlm.nih.gov/pubmed/12838606.

Maxwell, Colin S., and Magwene, Paul M. The Quick and the Dead: Microbial Demography at the Thermal Limit. In press.

Miller, Christine, Line E. Thomsen, Carina Gaggero, Ronen Mosseri, Hanne Ingmer, and Stanley N. Cohen. 2004. “SOS Response Induction by Beta-Lactams and Bacterial Defense against Antibiotic Lethality.” Science 305 (5690): 1629–31. doi:10.1126/science.1101630.

Mirelman, David, Gideon Altmann, and Yuval Eshdat. 1980. “Screening of Bacterial Isolates for Mannose-Specific Lectin Activity by Agglutination of Yeasts.” Journal of Clinical Microbiology 11 (4): 328–31.

Mortensen, Henrik D., Kitt Dupont, Lene Jespersen, and Nils Arneborg. 2007. “The Flo11p-Deficient Saccharomyces Cerevisiae Strain Background S288c Can Adhere to Plastic Surfaces.” Colloids and Surfaces B: Biointerfaces 60 (1): 131–34. doi:10.1016/j.colsurfb.2007.06.004.

67

Mortimer, Robert K. 2000. “Evolution and Variation of the Yeast (Saccharomyces) Genome.” Genome Research 10 (4): 403–9. doi:10.1101/gr.10.4.403.

Peng, X., J. Sun, D. Iserentant, C. Michiels, and H. Verachtert. 2001. “Flocculation and Coflocculation of Bacteria by Yeasts.” Applied Microbiology and Biotechnology 55 (6): 777–81. doi:10.1007/s002530000564.

Reynolds, Todd B., and Gerald R. Fink. 2001. “Bakers’ Yeast, a Model for Fungal Biofilm Formation.” Science 291 (5505): 878–81. doi:10.1126/science.291.5505.878.

Rigden, Daniel J., Luciane V. Mello, and Michael Y. Galperin. 2004. “The PA14 Domain, a Conserved All-β Domain in Bacterial Toxins, Enzymes, Adhesins and Signaling Molecules.” Trends in Biochemical Sciences 29 (7): 335–39. doi:10.1016/j.tibs.2004.05.002.

Rupp, Steffen, Eric Summers, Hsiu-Jung Lo, Hiten Madhani, and Gerald Fink. 1999. “MAP Kinase and cAMP Filamentation Signaling Pathways Converge on the Unusually Large Promoter of the Yeast FLO11 Gene.” The EMBO Journal 18 (5): 1257–69. doi:10.1093/emboj/18.5.1257.

Rutherford, Steven T., and Bonnie L. Bassler. 2012. “Bacterial Quorum Sensing: Its Role in Virulence and Possibilities for Its Control.” Cold Spring Harbor Perspectives in 2 (11): 1–25. doi:10.1101/cshperspect.a012427.

Schauder, Stephan and Bassler, Bonnie. 2001. “The Languages of Bacteria.” Genes and Development 15: 1468–80. doi:10.1101/gad.899601.attracting.

Schindelin, Johannes, Ignacio Arganda-Carreras, Erwin Frise, Verena Kaynig, Mark Longair, Tobias Pietzsch, Stephan Preibisch, et al. 2012. “Fiji: An Open-Source Platform for Biological-Image Analysis.” Nature Methods 9 (7): 676–82. doi:10.1038/nmeth.2019.

Schwartz, Monica A., and Hiten D. Madhani. 2004. “Principles of MAP Kinase Signaling Specificity in Saccharomyces Cerevisiae.” Annual Review of Genetics 38: 725–48. doi:10.1146/annurev.genet.39.073003.112634.

Smit, Gerrit, Marika H. Straver, Ben J. Lugtenberg, and Jan W. Kijne. 1992. “Flocculence of Saccharomyces Cerevisiae Cells Is Induced by Nutrient Limitation , with Cell Surface Hydrophobicity as a Major Determinant.” Applied and Environmental

68

Microbiology 58 (11): 3709–14.

Sperandio, Vanessa, Alfredo G. Torres, Jorge A. Giron, and James B. Kaper. 2001. “Quorum Sensing Is a Global Regulatory Mechanism in Enterohemorrhagic Escherichia Coli O157:H7.” Journal of Bacteriology 183 (17): 5187–97. doi:10.1128/JB.183.17.5187.

Strope, Pooja K., Daniel A. Skelly, Stanislav G. Kozmin, Gayathri Mahadevan, Eric A. Stone, Paul M. Magwene, Fred S. Dietrich, and John H. McCusker. 2015. “The 100- Genomes Strains, an S. Cerevisiae Resource That Illuminates Its Natural Phenotypic and Genotypic Variation and Emergence as an Opportunistic Pathogen.” Genome Research 125 (5): 762–74. doi:10.1101/gr.185538.114.

Surette, Michael G., and Bonnie L. Bassler. 1998. “Quorum Sensing in Escherichia Coli and Salmonella Typhimurium.” Proceedings of the National Academy of Sciences of the United States of America 95 (12): 7046–50. doi:10.1073/pnas.95.12.7046.

Taga, Michiko E., and Karina B. Xavier. 2011. “Methods for Analysis of Bacterial Autoinducer-2 Production.” Current Protocols in Microbiology, 1–15. doi:10.1002/9780471729259.mc01c01s23.

Verstrepen, K., G. Derdelinckx, H. Verachtert, and F. R. Delvaux. 2003. “Yeast Flocculation: What Brewers Should Know.” Applied Microbiology and Biotechnology 61 (3): 197–205. doi:10.1007/s00253-002-1200-8.

Verstrepen, Kevin J., and Frans M. Klis. 2006. “Flocculation, Adhesion and Biofilm Formation in Yeasts.” Molecular Microbiology 60: 5–15. doi:10.1111/j.1365- 2958.2006.05072.x.

Vlamakis, Hera, and Roberto Kolter. 2010. “Biofilms.” Cold Spring Harbor Perspectives in Biology 2 (a000398): 1–12.

Vyas, Valmik K., Sergei Kuchin, Cristin D. Berkey, and Marian Carlson. 2003. “Snf1 Kinases with Different β -Subunit Isoforms Play Distinct Roles in Regulating Haploid Invasive Growth Snf1 Kinases with Different ␤ -Subunit Isoforms Play Distinct Roles in Regulating Haploid Invasive Growth.” Molecular and Cellular Biology 23 (4): 1341–48. doi:10.1128/MCB.23.4.1341.

Wainright, Milton. 1997. “Pleiomorphism, a Forgotten Controversy.” Perspectives in Biology and Medicine 40 (3): 407–14. doi:10.1353/pbm.1997.0038.

69

Wainwright, M., L. T. Canham, K. Al-Wajeeh, and C. L. Reeves. 1999. “Morphological Changes (including Filamentation) in Escherichia Coli Grown under Starvation Conditions on Silicon Wafers and Other Surfaces.” Letters in Applied Microbiology 29 (Canham 1990): 224–27. doi:10.1046/j.1365-2672.1999.00602.x.

Wickham, Hadley. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York, 2009

Williams, Paul, Klaus Winzer, Weng C. Chan, and Miguel Camara. 2007. “Look Who’s Talking: Communication and Quorum Sensing in the Bacterial World.” Philosophical Transactions of the Royal Society B: Biological Sciences 362 (1483): 1119–34. doi:10.1098/rstb.2007.2039.

Wilson, W. James. 1905. “Pleomorphism, as Exhibited by Bacteria Grown on Media Containing Urea.” Pathological Laboratory of Queen’s College, Belfast.

Wisplinghoff, Hilmar, Tammy Bischoff, Sandra M. Tallent, Harald Seifert, Richard P. Wenzel, and Michael B. Edmond. 2004. “Nosocomial Bloodstream Infections in US Hospitals: Analysis of 24,179 Cases from a Prospective Nationwide Surveillance Study.” Clinical Infectious Diseases: An Official Publication of the Infectious Diseases Society of America 39 (3): 309–17. doi:10.1016/S1077-9108(08)70211-1.

Wolcott, R., J. W. Costerton, D. Raoult, and S. J. Cutler. 2013. “The Polymicrobial Nature of Biofilm Infection.” Clinical Microbiology and Infection 19 (2). European Society of Clinical Infectious Diseases: 107–12. doi:10.1111/j.1469-0691.2012.04001.x.

Wortinger, Mark A., Ellen M. Quardokus, and Yves V. Brun. 1998. “Morphological Adaptation and Inhibition of Cell Division during Stationary Phase in Caulobacter Crescentus.” Molecular Microbiology 29 (4): 963–73. doi:10.1046/j.1365- 2958.1998.00959.x.

Xavier, Karina B., and Bonnie L. Bassler. 2005. “Regulation of Uptake and Processing of the Quorum-Sensing Autoinducer AI-2 in Escherichia Coli.” Journal of Bacteriology 187 (1): 238–48. doi:10.1128/JB.187.1.238-248.2005.

Zara, Giacomo, Severino Zara, Claudia Pinna, Salvatore Marceddu, and Marilena Budroni. 2009. “FLO11 Gene Length and Transcriptional Level Affect Biofilm- Forming Ability of Wild Flor Strains of Saccharomyces Cerevisiae.” Microbiology 155 (12): 3838–46. doi:10.1099/mic.0.028738-0.

70

Zuo, Rongjun, Yoshifumi Hashimoto, Li Yang, William E. Bentley, and Thomas K. Wood. 2006. “Autoinducer 2 Controls Bio Lm Formation in.” Journal of Bacteriology 188 (1): 305–16. doi:10.1128/JB.188.1.305.

71