ORIGINAL PAPER

Functional roles for redox genes in ethanol sensitivityin Drosophila

Awoyemi A. Awofala & Jane A. Davies & Susan Jones

Received: 20 December 2011 /Revised: 23 February 2012 /Accepted: 28 February 2012 /Published online: 20 March 2012# Springer-Verlag 2012

Abstract Whilst the effects and associated costs of exces-sive alcohol consumption in the human population are ob-vious at one level, the roles played by genetic factors at themolecular level are still unclear. Drosophila melanogasterhas an alcohol response comparable to humans and is usedas a genetic model to study the functional roles of genesregulated in response to ethanol. In the current study, thebiological processes associated with behavioural responsesto acute alcohol exposure in Drosophila have been analysedusing whole genome expression profiling. Ethanol responsegenes differentially expressed (a) at a single time point (2 h)and (b) in a time series (0–4 h) were identified using micro-arrays. In addition, a subset of differentially expressed geneswas validated using behavioural sedation and recovery assays.The study shows that genes involved in redox processes,neuron development, and specific signalling and metabolicpathways (including glutathione metabolism) form part of theresponse to ethanol in Drosophila. Biological processes forthe regulation of oxidative stress are the common functional

denominator of many of the ethanol response genes identified.These upregulated genes work to rescue cells from oxidativestress and its consequences such as protein misfolding, apo-ptosis and ageing. In the current study, an enrichment ofDrosophila genes linked to ageing is observed for the firsttime. The functional genomics data revealed by such studiescan be used to predict transcription networks of ethanol re-sponse genes, but the future lies in mapping these networks tothe human population, with the ultimate aim of identifyinggenetic factors for alcohol use disorders.

Keywords Ethanol .Drosophila . Ageing .Microarray .

Oxidative stress . Glutathione metabolism

AbbreviationsDEG Differentially expressed geneWT Wild typeTS Time seriesSTP Single time pointMAPK Mitogen-activated protein kinase

Introduction

Whilst the effects and associated costs of excessive alcoholconsumption in the human population are obvious at onelevel, the roles played by genetic factors at the molecular levelare still unclear. The effects of alcohol are a result of complexbut coordinated molecular changes, and genetic factorsinfluence sensitivity and tolerance. Drosophila melanogasterhave an alcohol response comparable to humans and have beendeveloped as a genetic model (Guarnieri and Heberlein 2003).Combined genetic and behavioural studies onD. melanogasterhave shown that alcohol modifies a range of inhibitory andexcitatory neurotransmitters, including gamma-aminobutyric

Electronic supplementary material The online version of this article(doi:10.1007/s10142-012-0272-5) contains supplementary material,which is available to authorized users.

A. A. Awofala : J. A. Davies : S. JonesSchool of Life Sciences, University of Sussex,Falmer,Brighton BN1 9QG, UK

A. A. AwofalaDepartment of Biological Sciences,Tai Solarin University of Education,Ijebu-Ode, Ogun State, Nigeria

S. Jones (*)Information and Computer Science,The James Hutton Institute,Invergowrie,Dundee, Scotland DD2 5DA, UKe-mail: sue.jones@hutton.ac.uk

Funct Integr Genomics (2012) 12:305–315DOI 10.1007/s10142-012-0272-5

acid (GABA) (Paul 2006), glycine (Spanagel 2009), acetyl-choline (Spanagel 2009), serotonin (Lovinger 1999) andvoltage-gated calcium channels (Walter and Messing 1999).The signalling pathways in which these neurotransmittersfunction are mainly conserved in higher organisms, includinghumans. However, these studies do not give a clear picture ofhow signalling molecules influence alcohol sensitivity andtolerance at the molecular or systems biology level.

Additional studies quantifying transcriptional profiles inDrosophila have revealed that at the molecular level, thegenetic factors involved in the response to alcohol are notlimited to neurotransmitters (Morozova et al. 2006; Urizar etal. 2007; Kong et al. 2010). Such studies reveal that genesdifferentially expressed upon exposure to alcohol includethose involved in stress, immunity, olfaction and metabo-lism. However, a comparison of the data shows extremelylow overlap of differentially expressed genes (DEGs) be-tween studies (<28% for all pairwise comparisons and just<5% in a triple study comparison; Kong et al. 2010). Thishighlights the fact that the molecular picture of alcoholresponses is complex, and that further studies are requiredto reach a stable set of DEGs, and to establish how thesegenes map to biochemical pathways and ultimately pheno-types. The potential to link genetic factors to phenotypes inthe human population could lead to the identification ofgenetic markers for alcohol sensitivity and tolerance, andultimately alcohol use disorders such as alcoholism.

In the current study, the biological processes associatedwith behavioural responses to acute alcohol exposure in D.melanogaster are analysed using microarrays at a single timepoint (STP) and across a time series (TS). In addition, a subsetof DEGs is validated using behavioural assays on Drosophilawith ethanol-related mutant alleles. This study shows thatgenes involved in redox processes, and specific signallingand metabolic pathways, form part of the response to ethanolinDrosophila. An enrichment of genes involved in the ageingprocess was also observed. The discussion of the current studyfocuses, in part, on the links between redox genes regulated inresponse to the ethanol exposure oxidative stress mechanismand their links to ageing.

Methods

DEGs were identified using the Affymetrix microarray plat-form at (a) a STP (2 h) and (b) across a TS (0–4 h). The DEGswere compared with (a) two published STP studies(Morozova et al. 2006: Urizar et al. 2007) and (b) one pub-lished TS study (Kong et al. 2010). In addition, an intra-studycomparison was made between the sets of DEGs from the STPand TS in the current study. A subset of seven DEGs from theSTP analysis was validated in behavioural assays usingDrosophila with ethanol-related mutant alleles.

Fly strains and genetics

The general control strain used as wild type (WT) was isoge-nised on the second and third chromosomes (i.e. w+; Iso2C;Iso 3I) and was provided by Cahir O’Kane (University ofCambridge, Cambridge, UK). This strain has been reportedto behave similarly to the commonly used Canton-S stock in arange of behavioural tests (Sharma et al. 2005). The mutantstrains used were: the mbf12 and the control P[mbf1+]; mbf12

flies from S. Hirose (National Institute of Genetics, Mishima,Japan), hop25 (recessive lethal), hop27 (recessive lethal),Hsp8308445 and Hsp83e6A (recessive lethal) from BloomingtonDrosophila stock centre at Indiana University (Bloomington,Indiana, USA). The following generated ‘control’ stocks werealso used for assessing the effects of genetic background onflies’ ethanol response: Basc/+ (generated from a crossbetween hop25 or hop27 and the WT), Hsp8308445/TM3 (gener-ated from a cross between male and female of this stock) andTM6B/+ (generated from a cross betweenHsp83e6A/TM6B andtheWT). Two- to 5-day-old male or female flies were used forbehavioural testing.

Microarray methods

One hundred 2–5-day-old male flies from an isogenicOregon R strain were anaesthetised using ice and placedinto acrylic exposure boxes. After a 30-min recovery andacclimatisation period, they were then exposed to 15 min ofethanol vapour (300 ml 98% ethanol/0.41/min with 100 mlwater/0.21/min). After exposure, the treated flies wereplaced in 25-ml falcon tubes for a defined recovery periodof 0, 0.25, 0.5, 1, 2, 3 and 4 h in a standard environment(food containing vials kept at 25 °C). For each time treat-ment, three replicates of pooled flies were collected, exceptfor the 2-h treatment where four replicates were collected. Apopulation of control flies were exposed to water vapour for15 min (300 ml water 0.41/min) and placed in 25-ml falcontubes for a defined recovery period of 0, 0.25, 0.5, 1, 2, 3and 4 h in a standard environment (food containing vialskept at 25 °C). There were eight replicates collected for thecontrol populations, two for the 2-h time point and one for allthe other time points. Control and ethanol-treated flies werefrozen, and total RNA was extracted from the heads usingTrizol. This was used to generate biotin-labelled cRNA forhybridisation to GeneChip® Drosophila Genome 1 Array,using standard Affymetrix protocols.

Microarray data normalization and analysis

All microarray intensity data were adjusted and normalizedusing robust multi-array average (RMA) (Irizarry et al.2003) in the Bioconductor affy package implemented in R(Gautier et al. 2004; Gentleman et al. 2004). This adjusts the

306 Funct Integr Genomics (2012) 12:305–315

intensity data to account for variation and calculates expres-sion values in three steps: background adjustment, normal-ization and summarization (Huber et al. 2005). A total of 22treated (ethanol) GeneChips and 8 control (water) treatedGeneChips were analysed. Density plots of RMA expres-sion value distributions of all arrays were very similar withno outliers (data not shown).

Differential gene expression between treated and controlflies was assessed using an empirical Bayes approach in theBioconductor limma package implemented in R (Smyth2005). The Benjamini and Hochberg method (Benjaminiand Hochberg 1995) was implemented to adjust for multiplehypothesis testing. This controls the expected proportion offalsely rejected hypotheses defined as the false discoveryrate (FDR), and results in adjusted p values for each gene.Differential gene expression was analysed at both a STP andacross a TS. The 2-h time point was used as the STP aspreliminary analysis showed it gave the largest differentialgene expression values (data not shown). The 2-h STPanalysis used four replicate GeneChips for treated and eightpooled control replicates to identify DEGs at an adjusted pvalue of 0.05. The TS analysis used replicate GeneChips fortreated and control flies to analyse seven contrasts at timepoints of 0, 0.25, 0.5, 1, 2, 3 and 4 h using an adjusted p valueof 0.05. For the TS, four replicates for the 2-h treatment timepoint was used, three replicates were used for all other treat-ment time points and a pooled set of eight replicates used forthe controls. A set of pooled control replicates was used asthere were insufficient replicates at individual time points toachieve effective normalization. Hence, at each time point, thetreated replicates were normalized against the water controlreplicates. Functional classification and clustering of differen-tially expressed genes were conducted using the Database forAnnotation, Visualisation and Integrated Discovery (DAVID;Dennis et al. 2003). DAVID allows a module-centric approach

for functional analysis of gene lists including functional clas-sification and clustering using Gene Ontologies (GO) (TheGene Ontology Consortium 2000) and Kyoto Encyclopediaof Genes and Genomes (KEGG) database pathways(Kanehisa et al. 2012).

Behavioural assays

A subset of seven DEGs (ana, Axn, hiw, hop, hsp26, hsp83and mbf1) derived from the 2-h time point analysis wasvalidated by conducting sedation and recovery assays onflies carrying mutant alleles for these genes. The sevengenes were selected based on the microarray gene expres-sion data (all had adjusted p values of <1.0e−2), availabilityof fly stocks and suitability for behavioural genetic testing(Table 1). Suitability was based on stocks being viable foreither males or females and having normal (or close tonormal) wings and body size as in wild-type stocks. Inaddition, the selection was made based on the occurrenceof a human homologue and previous data implicating thegene or related genes in cellular processes in the centralnervous system. In total, populations with 12 mutant alleles(ana1, AxnEY10228, hiwND8, hiwEP1305, hiwEP1308, hop25,hop27, hsp26EY10556, hsp26KG02786, hsp83e6A, hsp8308445

and mbf12), corresponding to seven genes, were tested.

Sedation assay for ethanol sensitivity

The sedation assay previously described (Wen et al. 2005)was modified and used to test candidate genes for alcoholsensitivity. Twenty active and well-fed males (or females inthe case of stocks that produced non-viable males) wereused for each trial. These flies were selected under CO2

anaesthesia and allowed to recover for 24 h before use. A1-ml ethanol solution at 50% concentration was added to a

Table 1 Seven DEGs selectedfor experimental validationusing sedation and recoveryassays based on presence ofhuman homologues, fly stockavailability, previous dataimplicating the gene family incellular processes in the centralnervous system and mutantswere suitable for genetic testing

aFly stocks obtained throughpersonal communications

Symbol FlyBase no. Gene name Mutant alleles Foldchange

q value Bloomingtonstock no.

ana CG8084 Anachronism ana1 −1.59 8.48e−03 8927

axn CG7926 Axin axney10228 1.42 1.37e−02 17649

hiw CG32592 Highwire hiwep1305 1.24 4.62e−02 11420

hiwep1308 11421

hiwND8 –a

hop CG1594 Hopscotch hop25 1.68 1.26e−02 8494

hop27 8493

hsp26 CG4183 Heat shockprotein 26

hsp26ey10556 7.29 8.23e−03 20186

hsp26KG02786 132132

hsp83 CG1242 Heat shockprotein 83

hsp8308445 4.18 8.47e−03 11421

hsp83e6A 5695

mbf1 CG4143 Multiproteinbridging factor 1

mbf12 1.83 6.85e−03 –a

Funct Integr Genomics (2012) 12:305–315 307

180-ml plastic fly bottle. Flies were then transferred imme-diately into the bottle and the bottle sealed. The active fliesremained on the top inside the bottle, and the sedated fliesthat dropped to the bottom were counted at 6-min intervals.The mean sedation time (MST) was used to measure a fly’sresistance to the sedative effects of ethanol and was calcu-lated as shown in Eq 1.

Mean sedation time ¼P

xt� t

Nð1Þ

where

xt number of flies sedated at a given time (t)t time of sedationN the total number of flies sedated

Recovery assay for ethanol sensitivity

The recovery assay previously described (Wen et al. 2005)was modified and used to test the alcohol sensitivity ofselected genes. Twenty active flies were exposed to ethanolvapour for 12 min in a vial, to which 1 ml of 100% ethanolwas slowly added by soaking a cotton plug. After thisexposure, all flies were motionless at the bottom of the vial.Subsequently, the ethanol-soaked plug was then replacedwith an ethanol-free plug. The number of active flies wasthen counted at 3-min intervals as they recovered. Thepercentage of recovered flies at each time interval in boththe mutant and non-mutant strains was calculated. The meanrecovery time (MRT) was used as a measure of a fly’s abilityto recover from the sedative effects of ethanol and wascalculated as shown in Eq. 2.

Mean recovery time MRTð Þ ¼P

xt� t

Nð2Þ

xt number of flies recovered at a given time (t)t time of recoveryN total number of flies recovered

Results

DEGs were identified using the Affymetrix microarray plat-form (a) at a STP (2 h) and (b) across a TS (0–4 h). The DEGs(with both up- and downregulated genes combined) werecompared with (a) two published STP studies (Morozova etal. 2006; Urizar et al. 2007) and (b) one published TS study(Kong et al. 2010). In addition, an intra-study comparison wasmade between the sets of DEGs from the STP and TS in thecurrent study. A subset of seven DEGs from the STP analysis

was validated in behavioural assays using Drosophila withethanol-related mutant alleles.

Differentially expressed genes at a single (2 h) time point

At the 2-h time point, a total of 222 genes (144 upregulated and78 downregulated) were differentially expressed (adjustedp value <0.05). Significantly enriched functional groups wereidentified within this gene set using DAVID (Huang et al.2009) and were defined as groups with an enrichment scoreof >1.3 (equivalent to non-log scale 0.05). This revealed theDEGs had diverse functions including nucleotide binding,stress responses, protein folding and metabolism (Table 2).

Behavioural assays for validation of genes differentiallyexpressed at a single time point

Oligonucleotide microarrays are used for high-throughputscreening of changes in gene expression level, but the resultscannot be correlated simply with biological significance.Whilst the use of non-genetic approaches such as quantitativePCR may be useful to validate the microarray data, suchanalysis does not reveal the biological significance of expres-sion level changes. Thus, in the current study, the validation ofa subset of seven differentially expressed genes at the 2-h timepoint was conducted using sedation and recovery assays. Intotal, ten mutant alleles, corresponding to six genes, showedsignificant alterations in ethanol sensitivity in the sedationassay. Four mutant alleles, corresponding to two genes, (mbfl2,hiwEP1305 and hiwND8) showed enhanced sensitivity and six

Table 2 Function annotation clusters (derived using DAVID (Huanget al. 2009) enriched in the 222 DEGs from the 2-h time point analysis

Annotationcluster (2 h)

Biological function term Enrichmentscore

1 Nucleotide binding 4.82

2 Stress response 3.81

3 Unfolded protein binding 3.74

4 Transferase activity 3.46

5 Chaperone activity 3.46

6 Response to stimulus 3.40

7 Co factor binding 2.31

8 Oxidation reduction 2.11

9 Dehydrogenase activity 1.93

10 Lipid metabolism 1.92

11 Fatty acid metabolism 1.87

12 FAD binding 1.53

A single representative biological function term is shown for eachcluster (e.g. nucleotide binding has 13 subcategories, including ribo-nucleotide binding, purine ribonucleotide binding, adenyl ribonucleo-tide binding, etc.)

308 Funct Integr Genomics (2012) 12:305–315

alleles, corresponding to four genes (hop25/+, hop27/+,hsp8308445, hsp83e6A/+, ana1, AxnEY10228) showed reducedsensitivity. As an example, the MST and MRT are shown forthe mutant alleles corresponding to three genes (hop, hsp andmbfl) (Fig. 1).

Inter-study comparison of differentially expressed genesat a single time point

The study of Morozova et al. (2006) identified a total of 534differentially expressed discrete genes, and the study ofUrizar et al. (2007) identified a total of 644 genes. Adetailed comparison of the methods and analysis conductedin both these studies is included in Supplementary Table S1.The 222 DEGs (both up- and downregulated) at the 2-h timepoint in the current study were compared against DEG setsfrom these two previous studies. A comparison betweenDEG sets revealed 19 genes common to all three studies(Fig. 2), which is 8.6 % of the DEGs in the current 2-h timepoint study. Eighteen of the 19 genes common to all three studies showed the same direction of expression regulation

(Table 3), with 12 genes upregulated and 6 genes downregu-lated. The remaining gene, CG4716, showed downregulationin the Morozova and the current study, but upregulation in theUrizar study. The function of the 19 DEGs has been assignedusing Biological Process terms from the Gene Ontology (TheGO Consortium 2010) (Table 3). This assignment shows thatfive genes (Sodh2, Cyp12e1, CG18522, Cyp6a8 and CG1443)are predicted to function in oxidation–reduction processes.Three genes also function in response to stimuli (Kraken, 1(2)efl, Cyp6a8), whilst two (ry and Cy6a8) have been experimen-tally shown to be involved in specific metabolic processes andone (RfaBp) in signalling pathways.

Differentially expressed genes identified from a time seriesanalysis

The TS analysis from 0 to 4 h in seven time points identified1,965 DE with respect to one or more time points (q<0.05). Acomparison of gene expression variance using Levene’s test(Carroll and Schneider 1985) at α00.05 showed that there arestatistically significant variations in gene expression at all timepoints (p<0.001). Using DAVID (Huang et al. 2009), the1,965 DEGs were shown to be significantly enriched in genesinvolved in diverse functions including neuronal differentia-tion, protein folding, membrane organisation, transcriptionand signalling (Table 4). Functional clusters, identified at theGO gene level, such as protein folding and transcriptionregulation are in themselves very broad, and it is difficult touse them to establish the functional relationship betweengenes. It is more informative to identify biochemical pathwaysthat are enriched within the DEG dataset to show how geneswith different functions may occur within the same pathway.Hence, the 1,965 DEGs were mapped to KEGG pathway

0102030405060708090

WT

WT

(F)

hop[25

]/+ (F

)

hop[27

]/+ (F

)

Basc/+(F

)

hsp8

3[08

445]

(F)

hsp8

3[08

445]

/TM

3(F)

hsp8

3[e6

A]/+ (F

)

TM6B

/+(F

)

mbf1[2]

mbf1[+]

MS

T, m

in

* *

*

*

*

05

1015202530354045

WT

WT (F

)

hop[25

]/+ (F

)

hop[27

]/+ (F

)

Basc/+ (F

)

hsp8

3[08

445]

(F)

hsp8

3[08

445]

/TM3(

F)

hsp8

3[e6

A]/+ (F

)

TM6B

/+(F

)

mbf1[2]

mbf1[+]

MR

T, m

in

***

*

*

a

b

Fig. 1 Mean sedation time (MST) (a) and mean recovery time (MRT)for Drosphila populations with mutant alleles corresponding to threegenes, hop, hsp83 and mbf1. WTwild type; *P<0.001 for each mutantvs. control; bars represent standard errors of the mean; F indicatefemale flies and where not specified, male flies. Full genotype detailsin “Methods” section

Fig. 2 Venn diagram showing intersect between the number of DEGsfrom the 2-h time point analysis in the current study (2 h) with thestudy by Morozova et al. (2006) (Morozova) and the study by Urizar etal. (2007) (Urizar). Only 19 genes are shown to be differentiallyexpressed in all the three studies

Funct Integr Genomics (2012) 12:305–315 309

(Kanehisa et al. 2012) using DAVID (Huang et al. 2009)and significantly enriched pathways identified. The mappingshowed seven pathways were enriched (p<0.001). In order ofsignificance, these were glutathione metabolism (map00480),mitogen-activated protein kinase (MAPK) signalling(map04010), metabolism of xenobiotics by cytochromeP450 (map00980), endocytosis (map04144), drug metabo-lism–cytochrome p450 (map00982), Wnt signalling(map04310) and RNA degradation (map03018).

Glutathione metabolismwas themost significantly enrichedpathway, and the gene expression profiles of six glutathionetransferase genes (GstD1, GstD2, GstE1, GstE3, GstE7 andGstE8) present in the 1,965 DEG dataset are shown inFig. 3a–f. These genes show variable expression profiles, butall have peak expression at the 2-h time. The expressionprofiles of three differentially expressed transcription factors(cbt, sug and Sox14), all of which show earlier expression peaksbetween 0.5 and 1 h, are shown for comparison (Fig. 3g–i).

Inter-study comparison of differentially expressed genesfrom a TS analysis

The DEGs (both up- and downregulated) from the currentTS analysis were compared with the Kong time series

Table 3 Biological process terms (as defined in the Gene Ontologydatabase (GO Consortium 2003)) for 19 genes shown to be differen-tially expressed in three STP microarray studies (Morosova et al. 2006,Urizar et al. 2007 and the current study)

Gene symbol GO BP Regulation

CG4797 Transmembrane transport (p) ?

Sodh-2 Oxidation–reduction process (p) D

ry Arginine metabolic process DGlycerophospholipidmetabolic process

Purine base/pyrimidinemetabolic process

Tryptophan metabolic process

Oxidation–reduction process (p)

Cyp12e1 Oxidation–reduction process (p) D

CG13283 Proteolysis (p) D

HmgZ – D

RfaBp Smoothened signalling pathway DLipid transport

Wnt receptor signalling pathway

Cher Determination of adult life span UMitosis

Olfactory learning

Protein localisation

Germline ring canal formation

CG18522 Oxidation–reduction process (p) U

Cyp6a8 Insecticide metabolic process ULaureic acid metabolic process

Response to caffeine

Oxidation–reduction process (p)

CG1443 Oxidation–reduction process (p) U

Pgd Pentose phosphate shunt (p) U

l(2)efl Embryo development (p) UResponse to heat (p)

GstD1 – U

Kraken Response to toxin UDigestion

CG14207 – U

CG6908 – U

CG15784 – U

CG4716 – U

Genes are classified as up- (U) or downregulated (D) or undefined (?)where the direction of regulation differs between studies

Table 4 Function annotation clusters (derived using DAVID (Huanget al. 2009) that are enriched in the 1,965 DEGs from the TS analysis

Cluster Function Enrichment score

1 Neuron differentiation 5.40

2 Membrane organisation 4.32

3 Protein folding 4.22

4 Regulation of transcription 4.09

5 Actin binding 3.84

6 Chaperone activity 3.44

7 Cell motility 3.15

8 Photoreceptor activity 3.02

9 Morphogenesis 2.90

10 Organ development 2.72

11 Transferase activity 2.69

12 Cytoskeleton organisation 2.60

13 Tube development 2.57

14 Kinase activity 2.48

15 Adherens junction 2.15

16 Sensory perception 2.11

17 ATPase activity 2.06

18 GTPase activity 2.03

19 Nucleotide binding 2.03

20 Gland development 1.98

21 Stress response 1.90

22 Chromatin modification 1.82

23 Photoreceptor cell differentiation 1.80

24 Regulation of mRNA processing 1.62

25 Protein localisation 1.62

26 GTPase activator activity 1.54

27 Sex differentiation 1.51

28 Regulation of neuron differentiation 1.49

29 Microtubule cytoskeleton 1.49

30 Brain morphogenesis 1.47

Annotation clusters with enrichment score >1.3 are shown in decreasingorder

310 Funct Integr Genomics (2012) 12:305–315

analysis that showed 1,807 DEGs across all six time points(Kong et al. 2010). A detailed comparison of the methodsand analysis conducted in this and the current study isincluded in Supplementary Table S1. The comparisonrevealed 175 genes common to both datasets (Fig. 4a).These 175 DEGs were annotated using DAVID (Huang etal. 2009) to identify significantly (score >1.3) enrichedfunctional clusters (Fig. 3b). This revealed seven clusterswith functions including oxidation–reduction, NAD binding,stress response, transferase activity, metabolism and proteinfolding.

Intra-study comparison of single time point and times seriesanalysis

A comparison of the 1,965 DEGs from the TS analysis inthe current study with the 222 DEGs from the 2-h time pointanalysis in the current study revealed a set of 163 commongenes (Fig. 4a1). These 163 DEGs were annotated using theDAVID tool (Huang et al. 2009) to identify significantly(score >1.3) enriched functional clusters (Fig. 4a2). Thisrevealed ten clusters with functions related to nucleotidebinding, transferase activity, response to stimulus andprotein folding.

Discussion

Initial studies into the molecular effects of alcohol, usingDrosophila as a genetically tractable model, revealed neuro-transmitters including GABA, acetylcholine and serotoninto be part of the complex and coordinated changes that occur(Lovinger 1999; Walter and Messing 1999; Paul 2006). Theuse of microarray technology has permitted the analysisof changes in gene expression at the genome level, and suchstudies reveal more diverse genetic factors being involved inthe response to alcohol, including those linked to stress, proteinfolding and metabolism (Morozova et al. 2006; Urizar et al.2007; Kong et al. 2010)

However, microarray studies conducted to date only showsmall (<20 %) overlaps between DEG sets (Kong et al. 2010),and comparisons made in the current study also reflect this.Only 8.6 % of DEGs identified at the 2-h time point wereclassified as differentially expressed in the Morozova et al.(2006) and Urizar et al. (Urizar 2007) studies. Variations inDEG sets stem from differences in fly stocks, size of flypopulations and microarray hybridization protocols. Such ex-perimental variations are compounded by the use of differentstatistics for significance thresholds for DEGs. Low reproduc-ibility between DEG sets is especially evident when stringentp value thresholds are used (Shi et al. 2005), and an analysis of

a q=8.5e-6 b q=2.6e-5 c q=3.7e-4

d q=9.6e-7 e q=4.0e-5 f q=3.3e-4

g q = 3.2e-6 h q=7.6e-6 i q=1.2e-4

GstD1

-1.5

-1

-0.5

0

0.5

1

1.5

2

0 0.25 0.5 1 2 3 4Fo

ld C

han

ge

GstE3

-2

-1

0

1

2

3

4

0 0.25 0.5 1 2 3 4

Fo

ld C

han

ge

GstE8

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

0 0.25 0.5 1 2 3 4Fo

ld C

han

ge

GstE1

-2

-1

0

1

2

3

4

5

0 0.25 0.5 1 2 3 4

Fo

ld C

han

ge

GstE7

-2

-1

0

1

2

3

4

0 0.25 0.5 1 2 3 4

Fo

ld C

han

ge

GstD2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3

0 0.25 0.5 1 2 3 4

Fo

ld C

han

ge

cbt

-2

-1

0

1

2

3

4

5

6

0 0.25 0.5 1 2 3 4

Time, h

Fo

ld C

han

ge

sug

-2

-1

0

1

2

3

4

0 0.25 0.5 1 2 3 4

Time, h

Fo

ld C

han

ge

Sox14

-1.5

-1

-0.5

0

0.5

1

1.5

2

0 0.25 0.5 1 2 3 4

Time, h

Fo

ld C

han

ge

Fig. 3 Expression levels expressed as fold change for two functionalclasses of DEGs from the TS analysis. The fold change at each timepoint is calculated as relative to the reference control flies that were

exposed to water vapour. (a–f) glutathione transferases (GstD1, GstD2,GstE1, GstE3, GstE7, GstE8) and (g–i) transcription factors (cbt, sox14,sug). Adjusted p values (labelled as q values) are shown for each gene

Funct Integr Genomics (2012) 12:305–315 311

DEGs in human tumours concluded that thousands of expres-sion profile replicates are required to achieve a 50 % overlapbetween gene sets (Ein-Dor et al. 2006). In contrast, otherstudies suggest that gene sets that do not show significantoverlaps with others can comprise true DEGs if they have lowindividual FDRs (Zhang et al. 2008). Collectively, these studiesindicate that, whilst microarrays have the potential to revealnovel genetic factors involved in diverse traits, analysis meth-ods are still evolving. Recently, new statistical measures, whichaccount for coordinated changes in gene expression, have beenproposed to achieve consistency between DEG sets with smallnumbers of replicates (Zhang et al. 2009). In the current work,extensive comparisons of DEG sets from different studies haverevealed common functional themes and identified novel ge-netic factors in the ethanol response.

In this study, common functions shared by genes differen-tially expressed 2 h after exposure to ethanol include oxida-tion–reduction processes, response to stimulus/stress, proteinfolding, nucleotide binding and fatty acid and lipid metabolism(Tables 2 and 3, Fig. 4). A comparison of the TS and STP datarevealed, as expected, a greater number of DEGs and a greaterdiversity of functions from the TS, with eight times as manyDEGs identified compared to the STP. Additional functionalclasses identified from the TS include neuron differentiation,

transcription regulation, membrane organisation (Table 4) andageing (Fig. 4). In addition, DEG sets from the TS analysisshowed an enrichment of genes involved in the MAPK andWnt signalling pathways. This analysis also revealed an en-richment of genes involved in glutathione metabolism. Genesinvolved in stress responses (hsp26, hsp83) and signalling(hiw, axn and hop) were also observed from the behaviouralassays (Fig. 1). The stress response gene hsp26 has previouslybeen shown to be required for ethanol tolerance in Drosophila(Awofala et al. 2011). The validation of gene mbf1 highlightedthe importance of coordinated expression as this gene is atranscriptional co-activator involved in the regulation of oxi-dative stress responses. The functional context of a number ofethanol-regulated genes highlighted in the current study hasbeen reported previously (e.g. fatty acid metabolism (Narce etal. 2001; Morozova et al. 2006)). In contrast, an enrichment ofgenes linked to ageing has not been explicitly reported frommicroarray experiments in Drosophila previously. The subse-quent discussion will focus on DEGs involved in transcriptionregulation and signalling, the fundamental mechanisms ofmost molecular responses and oxidative stress that links etha-nol with the ageing process.

Transcription factors (TFs) upregulated early in response toethanol could be responsible for the regulation of other ethanol-

1b1a

b2

Cluster Function Score 1 Oxidation-reduction 7.07 2 NAD binding 3.88 3 Stress response 3.78 4 Transferase activity 2.85 5 Fatty acid metabolism 2.37 6 Ageing 1.97 7 Protein folding 1.63

b2

Cluster Function Score1 Nucleotide binding 4.422 Transferase activity 4.213 Response to stimulus 4.214 Protein folding 4.045 Chaperone activity 2.886 Unfolded protein binding 2.487 Dehydrogenase activity 2.418 Co-factor binding 1.939 FAD binding 1.8610 Ageing 1.63

Fig. 4 Inter- (a) and intra- (b) study DEG comparisons. a1 Venndiagram showing intersect between the number of DEGs from the TSanalysis in the current study (TS) and the study of Kong et al. (2010)(Kong). a2 Function annotation clusters (derived using DAVID (Huanget al. 2009) enriched in the 175 overlapping DEGs. b1 Venn diagram

showing the overlap between the number of DEGs from the TSanalysis in the current study (TS) with the 2-h time point analysis(2 h) in the current study. b2 Function annotation clusters enriched inthe 163 overlapping DEGs. Annotation clusters with enrichment score>1.3 are shown ordered by score

312 Funct Integr Genomics (2012) 12:305–315

regulated genes observed in the current study. Three TFs, cbt,sox14 and sug, had expression levels that peaked at between0.5 and 1 h after exposure to ethanol (Fig. 3). Cbt is known tobe involved in the development of the nervous system(Belacortu et al. 2011), sox14 is required for 20E signalling atthe onset of metamorphosis (Ritter and Beckstead 2010) andsug plays a role in determining adult life span (Landis et al.2003). The TF mbf1 also showed enhanced sensitivity uponethanol exposure (Fig. 1), and this co-activator functions topreserve redox-dependant AP1 activity during oxidative stress(Jindra et al. 2004). The identification of specific ethanol-regulated TFs linked to neuronal development and redox pro-cesses provide the starting point for the development of tran-scription networks for ethanol responses. Co-regulatedtranscription networks for a set of 21 Drosophila genes withfunctions in ethanol sensitivity have been predicted in a recentstudy (Morozova et al. 2011). This work showed genes in-volved in nervous system development were present in extend-ed networks of ethanol sensitivity and resistance genes. Manyother functional categories were also present, including genesinvolved in transcription and oxidative phosphorylation.Whilstthe specific TFs identified in the current microarray analysis arenot observed in the transcription networks (probably due to thelimited gene data sets used to generate them and differences inexperimental protocols), general functional categories, such asneurone development, are common to both.

Alongside transcription, molecular signalling is fundamen-tal to most cellular processes. Three signalling genes, hop,hsp83 and hiw, were regulated by ethanol in the current study.Hop encodes a non-receptor tyrosine kinase involved in theJAK/STAT signalling pathway, which transmits signals fromoutside to inside the cell (Binari and Perrimon 1994; Luo et al.1999). Hsp83 is involved in the Raf-mediated signalling path-way (van der Straten et al. 1997) and is the homologue ofHsp90 (Yue et al. 1999). This pathway is characterised by thetransmission of signals from outside to inside the cell, whichultimately leads to the phosphorylation and nuclear transloca-tion of MAPK. An increase in the transcription of Hsp90 incortical neurons exposed to ethanol has previously been ob-served in mouse (Pignataro et al. 2007). Hiw encodes a ubiq-uitin ligase that regulates synaptic development at theDrosophila neuromuscular junction by downregulating a sig-nalling protein that promotes synaptic growth (Collins et al.2006). The pathway analysis from the TS experiment alsoimplicated the involvement of the MAPK and the WNT sig-nalling pathways. The MAPK pathway represents the point ofconvergence of a number of signalling pathways (Aroor et al.2010), and has previously been implicated in diverse effects ofalcohol (Aroor and Shukla 2004). The WNT signalling path-way plays a central role in neurogenesis (Zhang et al. 2011),and the regulation of genes within the pathway could belinked to the apoptosis of neurons that results from ethanol-

induced oxidative stress in higher organisms (Ramachandranet al. 2003).

Molecular oxidative stress is the common functional denom-inator of many of the genes identified as ethanol-regulated inthe current study, including transcription factors and signallingproteins. Evidence of a role for redox genes in ethanolresponses (specifically those in the oxidative phosphorylationgene ontology class) has also been shown in predicted tran-scriptional networks associated with increased sensitivity toethanol (Morozova et al. 2011). Ethanol-induced oxidativestress has been reported widely, and is an imbalance in a cell’sredox reactions associated with increases in reactive oxygenspecies (ROS) or decreases in antioxidant defences such asglutathione (González et al. 2007). An increased productionof glutathione acts as an antioxidant, as in its reduced form itdonates a reducing equivalent to ROS. An ethanol-inducedredox imbalance would explain the regulation of genes in-volved in oxidation–reduction processes and glutathione me-tabolism. In the current study, six glutathione transferaseenzymes were upregulated (Fig. 3), a potential response toincreased ROS. Oxidative stress is also linked to an increasein misfolded proteins and ultimately apoptosis. The apoptosisof neurons has previously been observed in response to oxida-tive stress (Ramachandran et al. 2003; Björk et al. 2006). In thiscontext, the ethanol-induced regulation of genes involved inneurogenesis and protein folding, such as chaperones, would beexpected and was indeed observed.

Oxidative stress is one theory of how ageing occurs at themolecular level, with experimental evidence for populationswith increased oxidative stress resistance mechanisms havinglonger life spans (Bokov et al. 2004). In Drosophila, ethanoleffectively induces an additional oxidative stress burden, andhence, the regulation of genes linked to ageing (such as hsp26and Trxr-1) would be expected. Interesting, a recent study ofCaenorhabditis elegans showed that high concentrations ofethanol resulted in a shorten life span, but low concentrationsactually extended the life span (Yu et al. 2011). It is possiblethat upregulation of genes in antioxidant pathways such asglutathione biosynthesis can reduce ROS within cells at lowerlevels of ethanol exposure, but at higher levels cells in oxidativestress can no longer be rescued, and hence, ageing processesare accelerated.

In the current study, the biological processes associated withbehavioural responses to acute alcohol exposure inDrosophilahave been analysed using whole genome expression profiling.This study shows that genes involved in redox processes, andspecific signalling and metabolic pathways, form part of theresponse to ethanol in Drosophila. The regulation of oxidativestress is the common functional denominator of many of theethanol response genes identified. These upregulated geneswork to rescue cells from oxidative stress and its consequencessuch as protein misfolding, apoptosis and ageing. Whilst such

Funct Integr Genomics (2012) 12:305–315 313

data can be used to predict transcription networks of ethanolresponse genes in Drosophila (Morozova et al. 2011), thefuture lies in mapping these networks and associated interac-tions to the human population, with the ultimate aim of iden-tifying genetic factors for alcohol use disorders such asalcoholism.

References

Aroor A, Shukla S (2004) MAP kinase signalling in diverse effects ofethanol. Life Science 74:2339–2364

Aroor AR et al (2010) Differential changes in MAP kinases, histonemodifications, and liver injury in rats acutely treated with ethanol.Alcohol Clin Exp Res 34(9):1543–1551

Awofala AA, Jones S, Davies JA (2011) The heat shock protein 26gene is required for ethanol tolerance in Drosophila. J Exp Neuro-sci 5:31–44

Belacortu Y et al (2011) Expression of Drosophila Cabut during earlyembryogenesis, dorsal closure and nervous system development.Gene Expr Patterns 11(3–4):190–201

Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: apractical and powerful approach to multiple testing. J R Stat SocSer B (Methodol) 57:289–300

Binari R, Perrimon N (1994) Stripe-specific regulation of pair-rulegenes by hopscotch, a putative Jak family tyrosine kinase inDrosophila. Genes Dev 8(3):300–312

Björk K et al (2006) Glutathione-S-transferase expression in the brain:possible role in ethanol preference and longevity. FASEB J 20(11):1826–1835

Bokov A, Chaudhuri A, Richardson A (2004) The role of oxidativedamage and stress in aging. Mech of Ageing Dev 125(10–11):811–826

Carroll RJ, Schneider H (1985) Levene’s tests for equality of variances.Stat Prob Lett 3(4):191–194

Collins CA et al (2006) Highwire restrains synaptic growth by atten-uating a MAP kinase signal. Neuron 51(1):57–69

Dennis G, Jr, Sherman BT, Hosack DA, Yang J, Gao W et al (2003)DAVID: Database for Annotation, Visualization, and IntegratedDiscovery. Genome Biol 4:P3

Ein-Dor L, Zuk O, Domany E (2006) Thousands of samples are neededto generate a robust gene list for predicting outcome in cancer.Proc Natl Acad Sci U S A 103(15):5923–5928

Gautier L, Cope LM, Bolstad BM, Irizarry RA (2004) affy-Analysis ofAffymetrix GeneChip data at the probe level. Bioinformatics. 20(3):307–315

Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M et al(2004) Bioconductor: open software development for computa-tional biology and bioinformatics. Genome Biol 5:R80

González A, Pariente J, Salido GM (2007) Ethanol stimulates ROSgeneration by mitochondria through Ca2+ mobilization andincreases GFAP content in rat hippocampal astrocytes. BrainRes 1178:28–37

Guarnieri DJ, Heberlein U (2003) Drosophila melanogaster, agenetic model system for alcohol research. Int Rev Neurobiol54:199–228

Huang DW, Sherman BT, Lempicki RA (2009) Bioinformatics enrich-ment tools: paths toward the comprehensive functional analysis oflarge gene lists. Nucleic Acids Res 37(1):1–13

Huber W, Irizarry RA, Gentlemen R (2005) Preprocessing overview.In: Bioinformatics and Computational Biology Solutions using Rand Bioconductor, R. Gentleman, V. Carey, S. Dudoit, R. Irizarry,W. Huber (eds.), Springer, New York, pp 431–442

Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, et al. (2003)Summaries of Affymetrix GeneChip probe level data. NucleicAcids Res 31:e15

Jindra M et al (2004) Coactivator MBF1 preserves the redox-dependent AP-1 activity during oxidative stress in Drosophila.EMBO J 23(17):3538–3547

Kanehisa M et al (2012) KEGG for integration and interpretationof large-scale molecular data sets. Nucleic Acids Res 40:D109–14

Kong EC et al (2010) Ethanol-regulated genes that contribute toethanol sensitivity and rapid tolerance in Drosophila. AlcoholClin Exp Res 34(2):302–316

Landis GN, Bhole D, Tower J (2003) A search for doxycycline-dependent mutations that increase Drosophila melanogaster lifespan identifies the VhaSFD, Sugar baby, filamin, fwd and Cctlgenes. Genome Biol 4(2):R8

Lovinger DM (1999) 5-HT3 receptors and the neural actions of alcohols:an increasingly exciting topic. Neurochem Int 35(2):125–130

Luo H et al (1999) The Drosophila JAK kinase hopscotch is requiredfor multiple developmental processes in the eye. Dev Biol 213(2):432–441

Morozova TV, Anholt RR, Mackay TF (2006) Transcriptional responseto alcohol exposure in Drosophila melanogaster. Genome Biol 7(10):R95

Morozova TV, Mackay TFC, Anholt RRH (2011) Transcriptional net-works for alcohol sensitivity in Drosophila melanogaster. Genetics187(4):1193–1205

Narce M et al (2001) Effect of ethanol on polyunsaturated fatty acidbiosynthesis in hepatocytes from spontaneously hypertensive rats.Alcohol Clin Exp Res 25(8):1231–1237

Paul SM (2006) Alcohol-sensitive GABA receptors and alcohol antag-onists. Proc Natl Acad Sci U S A 103(22):8307–8308

Pignataro L et al (2007) Alcohol regulates gene expression in neurons viaactivation of heat shock factor 1. J Neurosci 27(47):12957–12966

Ramachandran V et al (2003) Ethanol-induced oxidative stress pre-cedes mitochondrially mediated apoptotic death of cultured fetalcortical neurons. J Neurosci Res 74(4):577–588

Ritter AR, Beckstead RB (2010) Sox14 is required for transcriptionaland developmental responses to 20-hydroxyecdysone at the onsetof Drosophila metamorphosis. Dev Dyn 239(10):2685–2694

Sharma P et al (2005) Isogenic autosomes to be applied in optimalscreening for novel mutants with viable phenotypes in Drosophilamelanogaster. J Neurogenet 19:57–85

Shi L et al (2005) Cross-platform comparability of microarraytechnology: intra-platform consistency and appropriate dataanalysis procedures are essential. BMC Bioinforma 6(Suppl2):S12

Smyth GK (2005) Limma: linear models for microarray data. In:Bioinformatics and Computational Biology Solutions using Rand Bioconductor, R. Gentleman, V. Carey, S. Dudoit, R. Irizarry,W. Huber (eds.), Springer, New York, pp 397–420

Spanagel R (2009) Alcoholism: a systems approach from molec-ular physiology to addictive behavior. Physiolological Rev 89(2):649–705

The Gene Ontology Consortium (2000) Gene Ontology: tool for theunification of biology. Nature Genet 25:25–29

The GO Consortium (2010) The Gene Ontology in 2010: extensions andrefinements. Nucleic Acids Res 38:D331–D335, Database issue

Urizar NL et al (2007) Drosophila homer is required in a small set ofneurons including the ellipsoid body for normal ethanol sensitivityand tolerance. J Neurosc 27(17):4541–4551

van der Straten A et al (1997) The heat shock protein 83 (Hsp83) isrequired for Raf-mediated signalling in Drosophila. EMBO J 16(8):1961–1969

Walter HJ, Messing RO (1999) Regulation of neuronal voltage-gatedcalcium channels by ethanol. Neurochem Int 35(2):95–101

314 Funct Integr Genomics (2012) 12:305–315

Wen T et al (2005) Drosophila neuropeptide F and its receptor, NPFR1,define a signaling pathway that acutely modulates alcohol sensi-tivity. Proc Natl Acad Sci U S A 102(6):2141–2146

Yu X et al (2011) Beneficial and harmful effects of alcohol exposure onCaenorhabditis elegans worms. Biochem Biophys Res Commun412(4):757–762

Yue L et al (1999) Genetic analysis of viable Hsp90 alleles reveals acritical role in Drosophila spermatogenesis. Genetics 151(3):1065–1079

Zhang M et al (2008) Apparently low reproducibility of true differen-tial expression discoveries in microarray studies. Bioinforma(Oxf, England) 24(18):2057–2063

Zhang M et al (2009) Evaluating reproducibility of differential expres-sion discoveries in microarray studies by considering correlatedmolecular changes. Bioinformat (Oxf, England) 25(13):1662–1668

Zhang L et al (2011) The Wnt/β-catenin signaling pathway in the adultneurogenesis. Eur J Neurosci 33(1):1–8

Funct Integr Genomics (2012) 12:305–315 315