1

nature | methods

eSGA: E. coli synthetic genetic array analysis

Gareth Butland, Mohan Babu, J Javier Díaz-Mejía, Fedyshyn Bohdana, Sadhna Phanse, Barbara Gold, Wenhong Yang,

Joyce Li, Alla G Gagarinova, Oxana Pogoutse, Hirotada Mori, Barry L Wanner, Henry Lo, Jas Wasniewski, Constantine

Christopolous, Mehrab Ali, Pascal Venn, Anahita Safavi-Naini, Natalie Sourour, Simone Caron, Ja-Yeon Choi, Ludovic

Laigle, Anaies Nazarians-Armavil, Avnish Deshpande, Sarah Joe, Kirill A Datsenko, Natsuko Yamamoto, Brenda J

Andrews, Charles Boone, Huiming Ding, Bilal Sheikh, Gabriel Moreno-Hagelseib, Jack F Greenblatt & Andrew Emili

Supplementary figures and text:

Supplementary Figure 1 Construction of query mutation in a Hfr Cavalli donor strain.

Supplementary Figure 2 Selective outgrowth steps at 1,536-colony density to produce consistent colony sizes.

Supplementary Figure 3 An eSGA plate image shows occasional cases where the Isolate-1 and Isolate-2 versions of a

deletion strain behave differently.

Supplementary Figure 4 Effect of linkage suppression on recombination efficiency.

Supplementary Figure 5 Reverse confirmation of synthetic genetic interactions using custom mini-arrays.

Supplementary Methods

Supplementary Results

Supplementary Discussion

Note: Supplementary Tables 1–4 are available on the Nature Methods website.

Homologous

recombination

Gene

replacement

Transformation

Into Hfr strain

Chromosomal ORF

Homologous recombination

45-nt homologous to target gene

20-nt homologous to CmR cassette

PCR

amplification

ΔabcA-CmR

SUPPLEMENTARY FIGURE 1

Supplementary Figure 1. Construction of query mutation in a Hfr Cavalli donor strain.

The query mutation in chromosome of an E. coli Hfr strain is constructed by transformation and replacement

of a desired locus with a chloramphenicol drug resistance (CmR) cassette using an in vivo recombination

approach based on PCR amplification2. The CmR cassette is generated by PCR using 20-nt forward and

reverse primers homologous to the 3’ end of the CmR marker (shown in white arrow heads), and 45-nt at 5’

end that are homologous to the sequences flanking the target gene intended for deletion (shown in colored

boxes).

a

SUPPLEMENTARY FIGURE 2

aidB

Donor

only

ybcJ

yacL

yhhP

rusA

Selection of double

mutants on LB with Kan and Cm

(excluding selective outgrowth step)

aidB

Donor

only

ybcJ

yacL

yhhP

rusA

Selection of double

mutants on LB with Kan and Cm

(including selective outgrowth step)

b

aidB

Donor

only

ybcJ

yacL

yhhP

rusA aidB

Donor

only

ybcJ

yacL

yhhP

rusA

Selection of double

mutants on LB with Kan and Cm

(excluding selective outgrowth step)

Selection of double

mutants on LB with Kan and Cm

(including selective outgrowth step)

Supplementary Figure 2. Selective outgrowth steps at 1,536-colony density produce consistent colony sizes.

The results of the selective outgrowth steps of ybcJ (a) and rusA (b) query deletion (marked with CmR) strains

conjugated into a custom designed array containing selected recipient deletion (marked with KanR) strains are shown

as plate images. The ybcJ (a) and rusA (b) query deletion strains is conjugated with recipient deletion strains containing

the following mutants (columns left to right: aidB, ybcJ, yacL, yhhP, and rusA). The conjugants are pinned onto the

LB plate with kanamycin and chloramphenicol for a first selection of 24 hr at 32 °C (left side image of the panels

‘a’ and ‘b’). A second selection (right side image of the panels ‘a’ and ‘b’) is performed by pinning the double mutants

from the first selection onto a second LB plate with kanamycin and chloramphenicol for 24 hr incubation at 32 °C.

More enhanced and uniform growth of colonies on the assay plates is clearly visible when an additional outgrowth step

is included. Donor only represents an internal negative control used in the assay. Each recipient deletion mutant strains

on the array are pinned in 8 columns.

SUPPLEMENTARY FIGURE 3

ΔiscS-CmR and

ΔnuoN-KanR

“Isolate-1”

ΔiscS-CmR and

ΔnuoN-KanR

“Isolate-2”

Supplementary Figure 3. An eSGA plate image shows occasional cases where the Isolate-1 and Isolate-2

versions of a deletion strain behave differently.

Each plate contains 384 different recipient single gene deletion mutants marked with KanR from the

KEIO collection that were used to screen against an ΔiscS-CmR query deletion mutant marked with CmR.

Each conjugant is pinned 4 times to a density of 1,536 and double mutants were selected on LB media

containing chloramphenicol and kanamycin antibiotics. The recipient deletion mutants marked with KanR

showing synthetic lethal or sick interactions against the ΔiscS-CmR query mutant are highlighted in black

line. Two independent versions of each strain were created in the KEIO collection; the four spots in a line on

a plate for a given mutant include two “Isolate-1” versions and two “Isolate-2” versions (shown in black

arrows). Occasionally, the “Isolate-1” and “Isolate-2” versions of a gene mutant do not behave identically in

an eSGA assay, indicating one of the strains must be defective. Based on our 39 genome-wide eSGA screens,

it appears that 195 “Isolate-1” and 224 “Isolate-2” strains behave abnormally, suggesting that they contain

genetic duplications or unknown compensatory lesions.

isc operon

fdx hscAhscB iscS iscRiscU

iscA

SUPPLEMENTARY FIGURE 4

Supplementary Figure 4. Effect of linkage suppression on recombination efficiency.

Illustrative example showing spurious results obtained when ∆iscU-CmR donor strain was conjugated with strains

bearing null mutations in and around the isc operon. We identified genes that fail to generate vigorously growing

viable recombinants. Essential no recombination (that is, few viable double mutants) was observed with the mutations

∆iscR-KanR, ∆iscS-KanR, ∆iscA-KanR, and ∆hscB-KanR, reduced recombination with ∆hscA-KanR, and slightly

impaired recombination with ∆fdx-KanR, which flank the iscU locus. Donor only represents an internal negative

control used in the assay.

SUPPLEMENTARY FIGURE 5

Supplementary Figure 5. Reverse confirmation of synthetic genetic interactions using custom mini-arrays.

The results of eSGA mini-array confirmation using ydhD query deletion strain are shown as plate images. The results

of double mutant growth of ydhD and pdxB query deletion strains in combination with recipient mutations present in the

ISC array (a) contains the following mutants (columns left to right: hcaC, yfhR, trmJ, iscR, iscS, iscU, iscA, hscB, hscA,

fdx, pbpC and a no recipient control), while the results of double mutant growth of ydhD and pdxB query deletion strains

in combination with recipient mutations present in the SUF array (b) contains the following mutants (columns left to

right: ydiN, ydiI, ydiH, sufA, sufB, sufC, sufD, sufS, sufE, ynhG, ydhW and a no recipient control (represented as ‘C’)).

The control images show the results of double mutant growth of a ybaS donor mutant in combination with recipient

mutations present within ISC or SUF arrays respectively. The apparent false positive interactions observed sporadically

in the ydhD-Suf array were possibly due to residual linkage, given the close proximity of the two loci (<30 kbp).

a

b

Supplementary Methods

Arraying recipient F- Kan

R E. coli mutant strains for genome-wide screens

The recipient F- Kan

R (resistance to kanamycin) KEIO

1 E. coli mutant strains arrayed in

ninety 96-well plates were transferred to twenty-four 384-well plates containing 100µl of

the Luria-Bertani (LB) medium supplemented with 50 µg/ml kanamycin grown overnight

at 32 ºC to an optical density (OD) of ~0.4 to 0.6 at 600nm by gently shaking the cultures

at 190 rpm. The overnight culture of each deletion mutant is supplemented with 15%

glycerol and stored at -80 ºC in 384-well plates prior to pinning. List of KEIO E. coli

deletion mutants used in the genome-wide screens are shown in Supplementary Table 1.

In each plate, we included “donor specific control or DSC” and “recipient specific control

or RSC” as an internal quality controls.

Construction of query deletion mutation in an Hfr Cavalli donor strain

For each genome-wide screen, a query gene deletion mutation is first constructed by

homologous recombination in an E. coli Hfr Cavalli donor strain bearing an integrated

temperature-inducible -Red high efficiency recombination system2. A temperature-

inducible λ Red cassette in strain DY3302 marked with Amp

R (resistance to ampicillin)

was moved by P1 transduction into Hfr Cavalli prior to the construction of query deletion

mutation. Transformants were selected on LB agar supplemented, where required, with

ampicillin (100 µg/ml) and chloramphenicol (12.5 µg/ml). Following were the four

critical steps used in the construction of query gene deletion mutation in Hfr Cavalli:

1. Preparation and generation of linear DNA cassette

Standard PCR conditions were used to amplify linear DNA fragments with Expand long

template PCR system (Roche). The CmR cassette (resistance to chloramphenicol) was

amplified from the plasmid pKD33 with forward (5’-

AGATTGCAGCATTACACGTCTT-3’) and reverse (5’-

GGCTGACATGGGAATTAGC-3’) primers. The PCR products were purified using

Qiaquick PCR purification kit (Qiagen). The amplified linear DNA is suspended in sterile

water, quantified by spectrophotometer and stored at -20 ºC. The linear DNA cassette is

designed to replace the target gene for deletion with CmR

marker. The cassette is

generated by PCR using 20-nt forward (5’- GCGTGTTACGGTGAAAACCT - 3’) and

reverse (5’-TCGTCGTGGTATTCACTCCA- 3’) primers homologous to the 3’ end of

the CmR

cassette, and 45-nt at 5’ end that are homologous to the sequences flanking the

target gene intended for deletion. Thus, the inner 20-nt sequence is common for all

forward and reverse knock out primers, while the 45-nt outermost primer sequence is

specific to the targeted gene. The PCR using these primers and a DNA template

containing the marker cassette generates a linear DNA product with the cassette flanked

by target homology. The generated linear cassette is stored in -20 ºC and used as template

for all subsequent PCRs. The PCR amplifications were carried out in 50µl reactions

containing 1x reaction buffer I, 200µM dNTPs, 2µM of forward and reverse knock out

primers, 3.75 U Taq DNA polymerase (Roche) and 0.5µl chloramphenicol template

(5ng/µl). Reactions were run for 30 cycles: 94 °C for 3 min; 30 cycles of 94 °C for 1 min,

56 °C for 1 min, 72 °C for 2 min; followed by 72 °C for 10 min. The amplified PCR

products were analyzed by running 1% agarose gel electrophoresis.

2. Electroporation and mutant selection

The Hfr Cavalli strain carrying the λ Red cassette is grown in 50 ml LB medium with

ampicillin respectively, at 32 ºC to an OD of ~0.4 to 0.6 at 600 nm. The electroporation

competent cells were prepared as described elsewhere2. Two microliter of the amplified

PCR product of CmR cassette was mixed with 50 µl of the competent cells and

electroporated in an ice-cold 0.2 cm cuvette at 2.5KV with 25µF and 200 , followed by

the addition of 1 ml SOC medium, to allow entry of the linear CmR

cassette. After

incubating the cells at 32 ºC for at least 1 hr with shaking at 220 rpm, ~100 µl of the

electroporated cell mixture was spread onto an agar LB plate containing 12.5 µg/ml of

chloramphenicol to select transformants at 32 ºC.

3. PCR confirmation of recombinants

In order to confirm the knock out mutants by PCR, genomic DNA was purified from two

random recombinant clones. The DNA is amplified with two sets of knock out

confirmation primers, and the size of the fragments confirmed using 1% agarose gel

electrophoresis. The first set consists of 20-nt flanking primer located 200 base pairs

upstream of the deleted region and one internal primer complementary to the CmR

cassette. This method ensures the position and location of the CmR

cassette and target

gene, respectively. The second set involves the same upstream probe, paired with another

20-nt flanking primer located 200 base pairs downstream of the deletion to verify the

correct mutation and to rule out the possibility that the selected strain is a rare diploid,

with one gene copy replaced by the cassette and the other copy adjacent to it. The list of

knockout and confirmation primer pair sequences used in the construction of the query

mutations in strains selected for the genome-wide screens are shown in Supplementary

Table 2.

PCR reactions were carried out in 50 µl reactions containing 1x reaction buffer I,

200µM dNTPs, 2µM of forward and reverse knock out confirmation primers, 3.75U Taq

DNA polymerase (Roche) and 1µl genomic DNA. Reactions were run for 30 cycles:

94 °C for 3 min; 30 cycles of 94 °C for 1 min, 56 °C for 1 min, 72 °C for 2min; followed

by 72 °C for 10 min. The amplified PCR products were confirmed using 1% agarose gel

electrophoresis.

4. Storage of confirmed query deletion mutants

Two randomly selected and successfully confirmed recombinant clones for each query

deletion mutants were stored in cryovials containing 1.5 ml LB medium supplemented

with 12.5 µg/ml chloramphenicol and 15% glycerol.

Construction of double mutants in E. coli using an automated strain handling

arraying procedure

Briefly, the process can be broken down into 4 separate sections, mutant array

preparation, conjugation, selection of double mutants on double antibiotics and

quantifying the colony sizes. A six-day replica plating method was used to construct

double mutants in a high throughput manner.

(A) First day, the frozen glycerol stock cultures of query deletion mutant created in an

Hfr strain with the target gene replacement with CmR were grown overnight at 32 ºC in

rich LB liquid media supplemented with 34 µg/ml of chloramphenicol prior to pinning

the cultures at 384 density.

(B) Second day, the overnight culture of the query deletion mutant, and the frozen

glycerol stock culture from an ordered array of recipient deletion mutant, marked with

KanR, were pinned onto a LB medium containing 34 µg/ml chloramphenicol and 50

µg/ml kanamycin, respectively at 384 density (24 column x 16 row)

(C) Third day, conjugation was carried out onto a LB plate by pinning the overnight

query and recipient deletion mutants at a 384 density.

(D) Fourth day, conjugants were subjected to first round of selection onto a LB medium

containing chloramphenicol and kanamycin antibiotics. On this day, two 384 colony

density conjugation plates are combined in duplicate onto a LB medium containing

chloramphenicol and kanamycin antibiotics to create 1,536 colony density plate. During

the development process, we did perform several experiments on a genome-wide scale in

which we selected the conjugant first in the presence of one drug (kanamycin or

chloramphenicol) and then in the presence of both antibiotics. We observed that the

selection of conjugants in one antibiotic (kanamycin or chloramphenicol), followed by

selection using both antibiotics, resulted in highly variable numbers of conjugants after

the first selection step. In the current eSGA screening process, the donor and recipient

cells are conjugated at a ratio of ~1:1. In these conditions, the variability observed in

pinning the conjugants first onto plates with a single drug was much greater than with our

standard selection process, to the extent that it was not possible to complete these screens

and obtain quantifiable genetic interaction data. For this reason, we decided to select

conjugants immediately with both drugs rather than first selecting the conjugants in the

presence of only one drug.

(E) Fifth day, the colonies from the first double drug selection plate is pinned onto a

second double drug selection plate. This additional selective outgrowth step deemed to be

advantageous for two reasons. First, it has been previously reported that having > 1 spot

to assay greatly increases the accuracy of growth measurements4 allowing the natural

measure of variation in the standard deviation. Second, it has been noted during the

development process that direct assay of the conjugation plate via double antibiotic

selection led to highly variable colony size distribution on the assay plates which would

be highly problematic for assaying via colony size. This is illustrated in Supplementary

Figure 2.

(F) Sixth day, the colonies from each double drug plate is photographed using a Kaiser

RS1 camera stand (product code no. 5510) and a digital camera (Canon Powershot A640,

10 Megapixels) with illumination from two Testrite 16 x 24 light boxes (Freestyle

Photographic supplies product #1624). The captured images were saved as jpeg files and

growth phenotype of the double mutants was quantitatively assayed using an in-house

automated image processing system originally devised for yeast4. In each step of the

above process, the plates were incubated for 24 hrs at 32 ºC.

Custom mini-array eSGA screening

A. Query mutant construction

Query mutants were constructed by replacing the target gene in Hfr Cavalli with a

chloramphenicol resistance marker, using the λ-Red system on pKD465, with the

following modifications. Transformation of electro-competent cells with PCR product

was performed using a BTX ECM630 (Harvard Apparatus, MA) and 2 mm gap cuvettes

with the following electroporation parameters: 2.5 kV, 200 ohms, 25 uF followed by 2-hr

recovery in 800 ul SOC without antibiotic at 32 °C, 200 rpm. Cells were pelleted and

plated on LB-Cm plates which were incubated overnight at 42 °C. Cultures of positive

clones were streaked on LB-Amp agar plates to check curing of plasmid (cultures re-

cured at 42 °C, if necessary). To confirm mutations, 2 to 4 colonies were selected and

grown overnight in 3 ml LB-Cm and genomic DNA purified (Wizard Genomic DNA

Purification kit, Promega, WI). Mutations are then verified by PCR amplification of

isolated genomic DNA with locus-specific primers 200 bp up- and downstream of target

gene and the resulting product checked for size and compared to the product of wild-type

genomic DNA amplification as a control. Antibiotics were used at the following

concentrations throughout, Ampicillin (Amp): 100 µg/ml; Chloramphenicol (Cm): 34

µg/ml and Kanamycin (Kan): 25 µg/ml.

B. Custom made recipient mutant mini-arrays

Recipient mutant arrays were made for 4 operon regions with mutants of the KEIO

collection1 in a 96-well format, 11 strains per plate, with a column of 8 wells for each

strain:

Suf array:

1 2 3 4 5 6 7 8 9 10 11 12

ydiN ydiI ydiH sufA sufB sufC sufD sufS sufE ynhG ydhW C

Isc array:

hcaC yfhR trmJ iscR iscS iscU iscA hscB hscA fdx pbpC C

Gnt array:

yhgF feoA feoB yhgA bioH gntX gntY gntT malQ malP yrbA C

Tru array:

purF ubiX argT dedD dedA truA usg pdxB yfcJ yfcL hemF C

These mini-arrays represent the viable deletion mutants present in the KEIO

collection around the isc, suf, gntY and truA loci, with the exception of yrbA and hemF,

which were added to the Gnt and Tru arrays, respectively to facilitate their screening.

After inoculation into 100 ul LB-Kan in a 200 ul 96-well plate and overnight growth (at

32 °C, 200 rpm) a replica on LB-Kan agar (Nunc Omnitray, NY) was made by pinning

with a 96-pin replicator (V&P Scientific, San Diego, CA), and Nunc Microwell and

Omnitray Copier frames, and incubated at 32 °C overnight.

Query deletion arrays were made with 2 donor strains of interest (rows A-C and

E-G) with row D empty and a control strain in row H, as illustrated in the following 96-

well format:

1 2 3 4 5 6 7 8 9 10 11 12

A iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS

B iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS

C iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS iscS

D

E iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU

F iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU

G iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU iscU

H ybaS ybaS ybaS ybaS ybaS ybaS ybaS ybaS ybaS ybaS ybaS ybaS

A replica was pinned on LB-Cm agar for the conjugation.

C. Conjugation and Selection

The solid donor and recipient plates are pinned onto a LB-agar conjugation plate which is

incubated for 24 hrs at 32 °C. This conjugation plate is pinned onto a LB-Kan-Cm agar

plate for a first selection of 48 hrs at 32 °C. A second selection is achieved by pinning the

first selection onto a second LB-Kan-Cm agar plate for 24-hr incubation. Plates were then

inspected manually and colonies compared to other donor/recipient combinations, the

control donor/recipient combinations and row D and column 12 as negative controls.

Supplementary Results

Determination of adequate conjugation time

To determine an adequate conjugation time, we conjugated Hfr Cavalli ∆rusA-CmR with

various recipients for 2, 4, 12 and 24 hr periods. The rusA gene is located at 12.3 min, just

beyond oriT, and so is one of the very last genes to be transferred. As shown in Figure A,

only a few viable conjugants were obtained for short mating periods (<12 hr), while 24 hr

conjugation supported efficient transgenesis. Consequently, we have used 24 hr of

conjugation for all subsequent screens.

Since E. coli is haploid, only one resistance cassette can occupy a given locus.

Indeed, as expected, very few self-conjugants were observed in the ∆rusA-CmR

and

∆rusA-KanR donor-recipient deletion combination (Fig. A). The residual self-double-

mutants sometimes observed presumably represent rare partial chromosomal duplication

events.

To further confirm the reliability of our conjugation and selection procedures, we

examined the genomic status of double mutants produced after 24 hr conjugative transfer

of two unrelated loci (∆aidB-CmR into a ∆yacL-Kan

R recipient). Genomic analyses of

single viable colonies confirmed appropriate replacement of both relevant loci in the

selected conjugants (Fig. B).

Quantitative Image analysis and generation of genetic interaction (S) score

Thirty-nine E. coli genome-wide screens are processed in batch scoring mode using

colony scorer program4. The calibrated software algorithm is executed in Java program

that identifies colonies arrayed in grid format and determines the fitness as a function of

growth rate. The raw colony sizes of the double mutants were normalized to each plate as

a whole and corrected for systematic biases such as plate edge effects, inter-plate

variation effects, column and row competition growth among neighboring colonies and

pinning defects were then analyzed statistically to take into account for replicate

reproducibility and deviations from the median colony sizes4. During the normalization

process the plates containing bad colony sizes were also filtered. There may however, be

cases in which we are unable to detect or quantify SSL interactions because one of the

mutations causes the E. coli colony to have an unusual morphological (e.g. spreading)

phenotype that masks or exaggerates the interaction.

The unbiased variance of relative errors (VARRE) between the replicate colony

sizes on each plate and a plate middle mean (PMM) value refereed as the mean colony

sizes on each plate4 were then computed to correct for unusual larger colony sizes

observed in the outermost rows and columns than the small or normal sized colonies in

the centre of the plate. Such kind of regression analysis is deemed to be necessary in

determining the minimum bound of the relative error in calculating the S score (Fig. C).

To reduce systematic errors, the effects of mutations in recipient strains were

normalized across the results of many unrelated screens, and the effect of each mutation

in a donor strain was taken into account by using the normalized median colony sizes of

all double mutants arising from the same donor strain as control or termed as a reference

set. The relationship between the unbiased variance of relative errors and the normalized

colony sizes from the reference set is shown in Figure D. The data suggests that larger or

smaller colony sizes tend to have larger relative variance, for which an adjustment of the

variances of double mutants was used in calculating the interaction score (S).

The normalized median colony growth sizes were then used to generate the

interaction score (S) for each pair of genes. These values reflected both our confidence in

the presence of genetic interactions as well as on the strengths of interactions. The raw

and normalized colony sizes, |Z scores|, and interaction (S) scores generated from 39

genome-wide screens are shown in Supplementary Table 4.

The interaction scores from 39 screens were then plotted to determine for a

normal distribution trend. Indeed, the S scores generated from the 39 query mutant strains

followed a normal distribution. However, we were very keen in knowing whether the

observed tail distribution in the experimental or actual dataset is due to significant

fraction of observed S scores or due to random chance. So, a randomized dataset was

created for each query mutant screen where the row and column coordinates and the raw

colony sizes of each recipient mutant were shuffled. The randomized dataset for each

query mutant screen is permuted 1,000 times. The genetic interaction (S) scores were

then calculated for each query mutant screen from the permuted randomized dataset. This

permutation testing revealed a large number of interactions in the experimental dataset

than in the randomized set in the tail of the distribution that represents aggravating

genetic interactions (Fig. 3c).

We also performed a second randomization to evaluate whether the observed

genetic interaction between pairs of gene from 39 genome-wide screens are due to true

genetic interaction or due to position effect of the genes on each plate. In order to

demonstrate this hypothesis, we randomized the actual datasets by scrambling the row

and column coordinates of each gene in a plate. During this randomization process, the

original raw colony sizes of each gene extracted from the colony software were retained.

The genetic interaction (S) scores were then calculated for both the actual and

randomized datasets (Fig. E). We clustered the actual and random datasets independently

by arranging the query mutants on the x-axis and recipient array mutants on the y-axis. As

expected, majority of the genetic interaction pattern, including the Suf and Isc genes

(Fig. E) between the actual and random data set showed similar patterns of genetic

interactions suggesting that the observed genetic interactions in the actual data set is true

and not due to position effect of the genes on each plate. As an example, a sub-matrix of

the actual and randomized data for the pairs of gene independently clustered by crossing

the 39 query deletion mutants with their respective recipient KEIO null mutant

collections (both “Isolate-1” and “Isolate-2” strain versions) is shown in Figure E.

Genetic interactions of the Isc genes outside of the Suf system

Our screens also revealed several Isc relationships outside of Suf (Supplementary

Discussion). The SSL interactions of members of the Isc system with hemF, which is

involved in heme biosynthesis, represent one such example. The genes hemF and hemN

encode enzymes catalyzing oxygen-dependent and -independent conversion of

coproporphyrinogen III to protoporphyrinogen IX, respectively. While HemN contains an

Fe-S cluster necessary for catalytic activity6, HemF is manganese-dependent

7. In this

case, the observed SSL interactions likely occur because reduced levels of holo-HemN

resulting from inactivation of the Isc machinery create a dependency on hemF for heme

production.

Validating the reliability of the novel genetic interactions

The authenticity and specificity of many of the high confidence novel genetic interactions

with |Z score| cutoff ≥ 4 (P < 0.0001) were confirmed by reanalyzing the donor-recipient

combinations using custom mini-array screens. We conjugated Isc donor mutations to

genes covering the truA genomic region (“tru array”), on which we also included hemF,

and the gntY locus (“gnt array”), on which we included yrbA. While all Isc pathway

components displayed varying levels of aggravating interactions with hemF, gntY and

yrbA, only iscS displayed a SSL interaction with truA (Fig. F), consistent with our full-

genome screens.

We additionally validated interactions which had |Z score| cutoff ≤ 4, slightly

below the very high confidence cut-off, indicating that there are likely numerous bona-

fide SSL interactions outside of the high-confidence dataset. Interactions to certain isc

genes, such as iscS with pdxB (dehydrogenase involved in pyridoxal 5-phosphate

biosynthesis) and its neighboring gene usg (predicted semi-aldehyde dehydrogenase) are

one such example (Fig. G). Clear functional links exist between iscS and pdxB, given that

pyridoxal 5-phosphate is the cofactor present in both IscS and SufS. The reliability of the

functional interactions was confirmed by conjugating Isc pathway mutations to genes

covering the usg-pdxB genomic region (“tru array”). As shown in Figure F, only iscS

displayed SSL interactions with usg and pdxB, consistent with the full genome eSGA

screens. The specificity of the iscS-pdxB was further validated by measuring reciprocal

genetic interactions by conjugating pdxB donor mutant against the deletion mutants

present in and around the isc, (“ISC array”) and suf (“SUF array”) loci (Supplementary

Fig. 5). The pdxB donor mutant strain displayed a specific SSL interaction with iscS and

not to other Isc pathway components or members of the Suf pathway, suggesting that iscS

and pdxB are functionally linked.

Supplementary Discussion

Factors led to the decision to use Hfr Cavalli strain

There were two major factors that led to the decision to use the Hfr Cavalli strain. First,

BW25113, the base strain for the KEIO recipient collection, is a 13-step descendent of

BD792, which is an F- strain derived from W1485 (F

+), which is a direct descendent of

K-12 EMG2 wild-type1. Looking at the lineage of BW25113 and its predecessors, we

were unable to find a known Hfr derivative. Second, during the initial development

process, we focused on determining the feasibility of Hfr mediated conjugative transfer

as a method for combining deletion mutations between strains. Several well-defined Hfr

strains (Hfr Cavalli, Hfr Hayes and Hfr 3000) were assessed for the ability to make donor

strain mutations using the method of Yu et al2, and their relative efficiencies in

conjugative DNA transfer of chromosomal markers and effects of position and relative

orientation of oriT. We were able to isolate substitution mutants with Hfr Cavalli with

much higher efficiency than with Hfr Hayes, and in trail conjugation experiments the

overall efficiency of transfer and the number of ex-conjugants observed was considerably

greater with Hfr Cavalli (data not shown). These observations highlighted the variability

of different Hfr backgrounds as efficient donor strains. Use of an isogenic strain would

have therefore required not only the isolation of a stable recombinant BW25113 (Hfr),

but the determination if it was a suitably efficient donor background. Given these factors,

the Hfr Cavalli strain was chosen as the donor background for the development process.

Potential problems associated with using non-isogenic strains

Although there are potential problems associated with using non-isogenic strains, we

believe that, in practice, these problems occur quite rarely for the following reasons.

First, potential problems associated with using the Hfr Cavalli donor should arise mainly

due to sequence differences between the Hfr Cavalli and the BW25113 recipient strains.

Hfr Cavalli is also a derivative of K-12 EMG2, but via a different lineage (W6). It has 6

known “mutations” (E. coli Genetic Stock Center CSGC# 4410;

http://cgsc.biology.yale.edu/Strain.php?ID=8448), but no complete genome sequence of

this strain is available, raising the possibility that other unrecorded mutations might exist

in this strain. None of these 6 known mutations exist in BW25113, hence it is possible

that, during mating, BW25113 will inherit one or more of these mutant loci via

homologous recombination. Second, the reverse situation is also possible. Several mutant

loci have been incorporated into BW25113 throughout its history, and indeed it is lac-,

ara- and rha

- amongst others. All of these loci are thought to be wild-type in Hfr Cavalli

and so there is the possibility that the recipient strain could revert to, for example, lac+

due to linked or independent recombination events. We are currently conducting lac

screening of ex-conjugants to determine the extent to which this phenomenon occurs.

While the above conversions or reversions are possible, we believe that it is

highly unlikely that they would notably affect the genetic interaction data obtained in this

study. Linkage data reported here suggests that a secondary mutation would have to be

within 30 kbp of the donor mutant locus to be considered linked (i.e. frequency of

inheritance greater than expected by independent inheritance). If not within ~30 kbp of

the donor mutant locus, secondary mutations leading to these conversions or reversions

should be inherited independently and not be affected by the selection process. For

example, if lac+ is not linked to a donor mutant, then its rate of inheritance in the

recipient would most likely be dependent on other factors (i.e. distance from oriT, nature

of recombination events occurring during mating). The variable here is the

recombination, which would most likely be different for every conjugation. Therefore,

even within one experiment, if we have multiple conjugation spots for the same donor-

recipient pair (as in mini-array experiments), we would expect some conjugants to go

lac+ while others remain lac-. However, we have good reason to think that such events

are rare. If these mutant alleles were often having an effect on the genetic interaction

results then we would expect to see rampant variability (i.e. every conjugation could be

different, and so we would have intra-and inter-experiment variation). This is something

which we haven’t observed in our genome-wide screens, where the reproducibility

between the replicate experiments on average was between 0.5 and 0.7. We therefore

strongly believe that use of a non-isogenic donor does not affect the genetic interactions

obtained in this study.

Factors considered for optimal mating conditions

While it may be true that the increased growth rate in rich LB medium might increase the

rate of spontaneous gene duplication, we considered several factors when deciding on

optimal mating conditions. First, the use of rich LB medium for the conjugation step

resulted in more consistent growth of all the recipient KEIO collection mutants in a

reasonable time frame. Cells grown in minimal medium not only required much longer

incubation times (approximately 2 weeks to complete a full genome screen in minimal

medium versus 6 days using rich medium), but also displayed more growth rate

variability among the different mutants present in the collection. Second, we reasoned

that this variability may lead to inconsistent conjugation when a consistent number of

donor cells are mated with varying numbers of recipient cells (depending on the viability

of each KEIO mutant in minimal medium). Third, while gene duplication can be a factor

in that we observed small numbers of self double mutants, we found that this was not a

vital issue with the two-round selection protocol we used in our current eSGA screening

procedure. Indeed, in order to observe more uniform growth and minimize stress

conditions, we incubated all cells at 32 ºC during the conjugation and selection steps. For

any specific eSGA experiments that require the use of minimal medium to study genetic

interactions for genes and pathways that are essential under specific growth conditions,

mating could be initially performed in rich medium and then the phenotype of the double

mutants could be scored on minimal medium supplemented with various additives such

as amino acids (methionine, cysteine), vitamins, and cofactors.

Iron-sulfur cluster biosynthesis systems

The major iron-sulfur cluster biosynthesis systems are encoded by the iscRSUA-hscBA-

fdx operon (known as the Isc system8) and the sufABCDSE operon (known as the Suf

system9). We chose these systems for our proof-of-principle experiments because these

systems were well supported with reports of synthetic lethality in the literature, and the

multiple components of each system would allow many double mutant combinations to

be tested for synthetic genetic interactions.

In the Isc system, a transcriptional repressor, IscR, binds DNA and represses

transcription of the operon in response to the presence of an Fe-S cluster in the protein10

.

The cysteine desulfurase, IscS, catalyses the conversion of cysteine to alanine along with

the removal of sulfide, which is trapped as a cysteine persulfide (-S-SH) on the active site

cysteine residue8. This sulfide is then transferred to the Fe-S scaffold protein IscU where,

in combination with ferrous (Fe2+

) iron, Fe-S clusters are formed transiently, prior to their

insertion into target apo-proteins11

. IscA has been proposed to play a role as an Fe-S

scaffold capable of binding Fe-S clusters, as well as a source of iron for cluster formation

due to its high affinity for ferrous iron12

. The remaining components of the Isc system are

the Hsp70 and Hsp40 protein folding chaperone homologs, HscA and HscB, and the

ferredoxin, Fdx. The HscA protein has been shown to bind the IscU-type scaffold

protein13

and may together with HscB, act to destabilize the IscU-[Fe-S] complex and

enable delivery of the cluster into its target apo-protein8.

An alternative Suf system for Fe-S biosynthesis contains many components

paralogous to those of the Isc system. SufSE is a cysteine desulfurase9, SufA is an IscA-

like protein14

and SufBCD, a predicted ATPase, has been proposed to stimulate SufS and

accept sulfur from SufE, suggesting it may act as a molecular chaperone akin to HscAB

(with IscU)15

. Importantly, it has been shown that, mutants in the Isc system are of

varying fitness, they are viable and, likewise, mutants in the Suf system display little

growth defect (SSL) 8, 9

. However, combinations of deletions of genes in the isc and suf

operons were previously shown to result in synthetic lethality8, 9

. This was directly

confirmed using iscU and sufC as donor mutations in crosses against custom mini-arrays

(Fig. 2b-c). Our observation that any suf gene deletion, with the exception of sufA, was

indeed non-viable in combination with an iscU deletion, along with our observation of

synthetic sickness of an iscU and sufA double mutant, is in agreement with recently

reported work14

. In these examples, the lack of synthetic lethality for iscU and sufA can

be explained by the ability of iscA to at least partially substitute for sufA.

More thorough analysis of Isc and Suf SSL interactions in genome-wide screens

revealed that most combinations do display aggravating interactions, iscA and sufA are

exceptions and display more subtle interactions with other components of the Suf and Isc

pathways respectively (i.e. the synthetic sick combination of iscU and sufA), but appear

synthetic lethal with each other (Fig. 4b-c ). These data broadly support the idea of sufA

and iscA partial complementation; however several SSL interactions including sufA and

fdx and iscA and sufB require further investigation in order to understand their functional

basis.

Interactions with essential genes

Although aggravating genetic interactions among non-essential genes are usually

orthogonal to protein-protein interactions16, 17

, we detected some strong SSL interactions

involving essential protein complexes which overlap with protein-protein interactions.

For example, greA and greB, which interact physically with RNAP, were synthetic lethal

with rpoA-SPA, which encodes the essential subunit of RNAP. GreA and GreB

reactivate arrested, backtracked RNAP by stimulating endonucleolytic cleavage of

nascent RNA by the RNAP active centre to generate a new 3’-end18

. Recently, an in vitro

study has shown that Gre factors aid progression from abortive initiation to elongation19

.

Therefore, the observed SSL interactions of greA and greB with rpoA-SPA suggest that

active greA and greB may be important in vivo for progression from initiation to

elongation.

Another example is the SSL interaction between rpoS, which encodes the

stationary phase sigma factor, S, and cysB, a regulator that binds to RNAP

20. Since cysB

and rpoS both influence arginine decarboxylase (aidA) expression via the arginine acid

resistance system21

, we speculate there might be insufficient activation in the absence of

these two factors, resulting in a fitness defect.

Analogous approach could be used to look at genetic interactions in other bacterial

species

Indeed, E. coli has been used extensively to transmit non-replicating “suicide” plasmids

by conjugal mating to a variety of bacterial species, including Paracoccus denitrificans22

.

Using the current approach, a similar high-throughput conjugative system could therefore

be used to transfer suicide plasmids (which contain fragments of the DNA of the recipient

strain as a genomic library, individually interrupted at a desired locus with a selectable

marker, with enough flanking DNA on each side for RecA-mediated recombination) from

E. coli into an ordered array of deletion mutants of the desired host species for which a

complete deletion collection is available (e.g., Bacillus subtilis23

, Acinetobacter baylyi

ADP124

). Such a system may even be more efficient than the current system, as one

would be transferring only a plasmid containing a relatively small insert (< 10 kbp) of

homologous DNA, and not a potentially large section of chromosomal DNA. Transfer

efficiency should therefore be high and there should be less chance of unselected

recombination effects. Thus, the availability of recipient gene deletion mutant

collections23, 24

and E. coli shuttle vectors25

should allow for efficient double mutant

construction and characterization for other prokaryotes.

Supplementary References

1. Baba, T. et al. Construction of Escherichia coli K-12 in-frame, single-gene

knockout mutants: the KEIO collection. Mol. Syst. Biol. 2, 2006.0008 (2006).

2. Yu, D. et al. An efficient recombination system for chromosome engineering in

Escherichia coli. PNAS 97, 5978-83 (2000).

3. Zeghouf, M. et al. Sequential Peptide Affinity (SPA) system for the identification

of mammalian and bacterial protein complexes. J. Proteome Res. 3, 463-468

(2004).

4. Collins, S. R., Schuldiner, M., Krogan, N. J. & Weissman, J. S. A strategy for

extracting and analyzing large-scale quantitative epistatic interaction data.

Genome Biol. 7, R63 (2006).

5. Datsenko, K.A. & Wanner, B.L. One-step inactivation of chromosomal genes in

Escherichia coli K-12 using PCR products. Proc. Natl. Acad. Sci. U S A 97, 6640-

6645 (2000).

6. Layer, G. et al. Radical S-adenosylmethionine enzyme coproporphyrinogen III

oxidase HemN. J. Biol. Chem. 280, 29038-29046 (2005).

7. Breckau, D., Mahlitz., E., Sauerwald, A., Layer, G. & Jahn, D. Oxygen-dependent

coproporphyrinogen III oxidase (HemF) from Escherichia coli is stimulated by

Manganese. J. Biol. Chem. 278, 46625-46631 (2003).

8. Tokumoto, U. & Takahashi, Y. Genetic analysis of the isc operon in Escherichia

coli involved in the biogenesis of cellular iron-sulfur proteins. J. Biochem. 130,

63-71 (2001).

9. Outten, F.W., Djaman, O. & Storz, G. A suf operon requirement for Fe-S cluster

assembly during iron starvation in Escherichia coli. Mol. Microbiol. 52, 861–872

(2004).

10. Schwartz, C.J. et al. IscR, an Fe-S cluster -containing transcription factor,

represses expression of Escherichia coli genes encoding Fe-S cluster assembly

proteins. Proc. Natl. Acad. Sci. U S A 98, 14895-14900 (2001).

11. Hoff, K.G., Silberg, J.J. & Vickery, L.E. Interaction of the iron-sulfur cluster

assembly protein IscU with the Hsc66/Hsc20 molecular chaperone system of

Escherichia coli. Proc. Natl. Acad. Sci. U S A 97, 7790-7795 (2000).

12. Takahashi, Y. & Nakamura, M. Functional assignment of the ORF2-iscS-iscU-

iscA-hscB-hscA-fdx-ORF3 gene cluster involved in the assembly of Fe-S clusters

in Escherichia coli. J. Biochem. 126, 917-926 (1999).

13. Hoff, K.G., Ta, D.T., Tapley, T.L., Silberg, J.J. & Vickery, L.E. Hsc66 substrate

specificity is directed toward a discrete region of the iron-sulfur cluster template

protein IscU. J. Biol. Chem. 277, 27353-27359 (2002).

14. Lu, J., Yang, J., Tan, G. & Ding, H. Complementary roles of SufA and IscA in the

biogenesis of iron-sulfur clusters in Escherichia coli. J. Biochem. 409, 535-543

(2008).

15. Vickery, L.E. & Cupp-Vickery, J.R. Molecular chaperones HscA/Ssq1 and

HscB/Jac1 and their roles in iron-sulfur protein maturation. Cri. Rev. Biochem.

Mol. Biol. 42, 95-111 (2007).

16. Boone, C., Bussey, H. & Andrews, B.J. Exploring genetic interactions and

networks with yeast. Nat. Rev. Genet. 8, 437-449 (2007).

17. Beyer, A., Bandyopadhyay, S. & Ideker, T. Integrating physical and genetic

maps: from genomes to interaction networks. Nat. Rev. Genet. 8, 699-710 (2007).

18. Komissarova, N. & Kashlev, M. Arrest of transcription: E.coli RNA polymerase

translocates backward leaving the 3' end of the RNA intact and extruded. Proc.

Natl. Acad. Sci.U.S.A. 94, 1755-1760 (1997).

19. Stepanova, E. et al. Analysis of promoter targets for Escherichia coli transcription

elongation factor GreA in vivo and in vitro. J. Bacteriol. 189, 8772-8785 (2007).

20. Lochowska, A., Nowicka, R.I., Plochocka, D. & Hryniewicz, M.M. Functional

dissection of the LysR-type CysB transcriptional regulator: regions important for

DNA binding, inducer response, oligomerization and positive control. J. Biol.

Chem. 276, 2098-2107 (2001).

21. Dong, T., Joyce, C. & Schellhorn, H.E. The role of RpoS in bacterial adaptation.

Bacterial Physiology: A Molecular Approach (ed.W.El-Sharoud) 313-337

(Springer-Verlag, Berlin, Heidelberg, 2008).

22. Baker, S.C. et al. Molecular genetics of the genus Paracoccus: Metabolically

versatile bacteria with bioenergetic flexibility. Micorbiol Mol Biol Rev. 62, 1046-

1078 (1998).

23. Kobayashi, K., Kondo, M., Fukuda, H., Nishimura, M. & Ohta, H. Essential

Bacillus subtilis genes. Proc. Natl. Acad. Sci.U.S.A 100, 4678-4683 (2003).

24. de Berardinis, V. et al. A complete collection of single-gene deletion mutants of

Acinetobacter baylyi ADP1. Mol. Syst. Biol. 4, 174 (2008).

25. Metzgar, D. et al. Acinetobacter sp. ADP1: an ideal model organism for genetic

analysis and genome engineering. Nucleic Acids Res. 32, 5780-5790 (2004).

a

aidB

Donor

only

ybcJ

yacL

yhhP

rusAaidB

Donor

only

ybcJ

yacL

yhhP

rusA

aidB

Donor

only

ybcJ

yacL

yhhP

rusA aidB

Donor

only

ybcJ

yacL

yhhP

rusA

Figure A. Determination of adequate time frame for conjugation.

The results of the time course interval experiment performed independently by allowing the donor deletion mutant rusA

(one of the very last gene to be transferred) to conjugate with custom designed array containing various selected recipient

deletion strains (columns left to right: aidB, ybcJ, yacL, yhhP, and rusA) for 2 (a), 4 (b), 12 (c) and 24 (d) hr periods are

shown as plate images. After a specified time course interval of conjugation, the conjugants are pinned onto the LB plate

with double antibiotics (kanamycin and chloramphenicol) for a first selection of 24 hr at 32 C. A second selection is

performed by pinning the double mutants from the first selection plate onto a second LB plate containing the double

antibiotics (kanamycin and chloramphenicol) for 24 hr incubation at 32 C. “Donor only” represents a negative control

with no recipient strain. Each recipient deletion mutant strains on the array are pinned in 8 columns.

FIGURE A

b

c d

FIGURE B

Figure B. Genomic analysis to confirm the knock outs in viable conjugants of two unrelated loci.

The genomic analysis of double mutants produced after 24 hr conjugative transfer tested using two unrelated

loci namely ∆aidB-CmR as a donor deletion strain and ∆yacL-KanR as a recipient deletion strain. The CmR or KanR

cassette is generated by PCR using 20-nt forward and reverse primers homologous to the 3’ end of the CmR or KanR

marker, and 45-nt at 5’ end that are homologous to the sequences flanking the target gene intended for deletion.

The sizes of the CmR and KanR cassettes , and aidB and yacL loci in base pairs are shown in parentheses.

Strains in which the target gene has been replaced by CmR or KanR cassette resulted in a product of 1,400 bp and

1,900 bp respectively, whereas false positive isolates containing the wild-type copy of the target gene resulted in a

product equal to the size of the gene plus 400 bp.

Lanes from left to right: M. DNA marker; Lane 1. ∆aidB-CmR query deletion strain amplified with yacL knock out

confirmation primers; Lane 2. ∆yacL-KanR recipient deletion strain amplified with yacL knock out confirmation

primers; Lane 3. ∆aidB-CmR query deletion strain amplified with aidB knock out confirmation primers;

Lane 4. ∆yacL-KanR recipient deletion strain amplified with aidB knock out confirmation primers;

Lanes 5 and 7 ; and 6 and 8 represent double mutants (∆aidB-CmR * ∆yacL-KanR) amplified with aidB and yacL

knock out confirmation primers, respectively. The molecular weights in bp are shown with an arrow head.

The lanes 1 to 4 serve as controls and show that donor and recipient is mutant for aidB and yacL, respectively

but wild type at the second locus. Lanes 5 to 8 show the same PCR amplifications using conjugant genomic

DNA and confirm that PCR products corresponding to both mutant loci are present in both isolates.

FIGURE C

Figure C. Regression analysis to determine the minimum bound of relative error in calculating the interaction

(S) score.

(a) Relationship between the unbiased variance of relative error and the plate-middle-mean of the eSGA plate. Blue

data points represent raw colony sizes and red data points represent the regression line of the raw colony sizes.

(b) Frequency distribution of the plate-middle-mean calculated for all double mutant colony sizes generated from

39 genome-wide screens.

(c) Regression (red line) between the unbiased variance of relative errors and the plate-middle-mean for all double mutant

colonies in each plate calculated from the weighted variance of the relative errors (blue dots).

a

b

c

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0 0.5 1 1.5 2 2.5

Un

bia

sed

va

ria

nce

of r

an

do

m

err

or

( V

AR

RE

)

Normalized colony Sizes

FIGURE D

Figure D. Relationship between the unbiased variance of relative random errors and the normalized colony sizes

from the reference set. The reference set is indicated as a control data set where the effect of each mutation in a query

deletion strain is taken into account by using the normalized median colony sizes of all double mutants arising from the

same donor strain.

a

b

Actual data

Randomized data

FIGURE E

Figure E. Evaluation of the genetic interaction between gene pairs from 39 genome-wide screens to determine

whether the observed genetic interaction is true or due to position effect of the genes on each plate.

Sub-matrix of S scores calculated for the actual (a) and randomized (b) dataset by conjugating 39 query deletion

mutants shown on the vertical axis with their respective recipient KEIO null mutants (both “Isolate-1” and “Isolate -2”

strain versions) on the horizontal axis. Closely linked gene pair can be clearly identified along the diagonal line. The

label “RAND” after the gene name represents randomization. The red color represents an aggravating (significantly

(P < 0.0001) negative S score) interaction, green color represents an alleviating (significantly (P < 0.0001) positive

S score) interaction and black color indicates absence of interaction.

FIGURE F

Figure F. Confirmation of synthetic genetic interactions using custom mini-arrays.

The eSGA mini-array confirmation results on using query deletion strains are shown as plate images. The results of

double mutant growth of iscS and iscU (a); iscA and hscA (c); and hscB and fdx (e) query deletion strains in combination

with recipient mutations present in the TRU array contains the following mutants (columns left to right: purF, ubiX,

argT, dedD, dedA, truA, usg, pdxB, yfcJ, yfcL, hemF and a no recipient control (represented as ‘C’)). The double mutant

growth of iscS and iscU (b); iscA and hscA (d); and hscB and fdx (f) query deletion strains in combination with recipient

mutations present in the GNT array contains the following mutants (columns left to right: yhgF, feoA, feoB, yhgA, bioH,

gntX, gntY, gntT, malQ, malP, yrbA and a no recipient control). The control images show the results of double mutant

growth of a sufC donor mutant in combination with recipient mutations present within TRU or GNT arrays respectively.

c

a e

d

bf

pdxB-usg-truA locus

IscS

ISC

System

gntY

yrbA

Protoporphyrinogen IX

Heme Biosynthesis

HemN

[FeS]HemF

Mn2+

Coproporphyrinogen III

ygfZ

FIGURE G

Figure G. Novel genetic interactions of iscS and the Isc system.

Novel genetic interactions uncovered in the eSGA screens associated with the Isc system are shown. Interactions

between hemF, gntY, ygfZ, yrbA and the components of the Isc system are indicated by double black lines. The

interaction between pdxB and truA, members of the pdxB-usg-truA locus, and iscS alone are indicated by a single

black line.