+ All Categories
Home > Documents > Blood genomic responses differ after stroke, seizures, hypoglycemia, and hypoxia: Blood genomic...

Blood genomic responses differ after stroke, seizures, hypoglycemia, and hypoxia: Blood genomic...

Date post: 31-Jan-2023
Category:
Upload: independent
View: 0 times
Download: 0 times
Share this document with a friend
9
Blood Genomic Responses Differ After Stroke, Seizures, Hypoglycemia, and Hypoxia: Blood Genomic Fingerprints of Disease Yang Tang, MD, 1 Aigang Lu, MD, 1 Bruce J. Aronow, PhD, 2 and Frank R. Sharp, MD 2 Using microarray technology, we investigated whether the gene expression profile in white blood cells could be used as a fingerprint of different disease states. Adult rats were subjected to ischemic strokes, hemorrhagic strokes, sham sur- geries, kainate-induced seizures, hypoxia, or insulin-induced hypoglycemia, and compared with controls. The white blood cell RNA expression patterns were assessed 24 hours later using oligonucleotide microarrays. Results showed that many genes were upregulated or downregulated at least twofold in white blood cells after each experimental condition. Blood genomic response patterns were different for each condition. These results demonstrate the potential of blood gene expression profiling for diagnostic, mechanistic, and therapeutic assessment of a wide variety of disease states. Ann Neurol 2001;50:699 –707 With the rapid advance of DNA microarray technol- ogy and sequencing of rodent and human genomes nearing completion, it is possible to examine the ex- pression of thousands of genes in a single sample of RNA derived from various tissues or organisms. 1,2 Mi- croarrays have been used to describe genomic responses to a variety of different stimuli, including serum stim- ulation, 3 iron, 4 glucose deprivation, 5 and hypoxia. 6 Disease-related gene expression patterns are being described, including genes specific for inflammatory disease, 7 evidence of toxic exposure, 8 and multiple scle- rosis. 9 Gene expression patterns in tumor tissue are dif- ferent from those in the host, 10 as well as different pat- terns in different tumor types. 11,12 Genomic responses in tumor cells from patients with acute myeloid leuke- mia differ from responses in acute lymphoblastic leu- kemia. 13 Of note has been the finding that even nor- mal peripheral white blood cells demonstrate different gene expression patterns when stimulated with differ- ent agents. 14,15 These latter findings prompted us to wonder whether the gene expression patterns in white blood cells would be distinct in different disease states, and whether the blood gene expression profile could serve as a fingerprint for various disease states. We chose to examine the blood gene expression profile because blood is the most readily accessible tissue in patients. Blood gene expression profiles might make it possible to diagnose many medical and toxic conditions that are currently difficult or impossible to assess. To begin to address this, we subjected adult rats to various brain- specific or systemic stimuli, including ischemic strokes, hemorrhagic strokes, sham surgeries, kainate-induced seizures, hypoxia, and insulin-induced hypoglycemia, and compared them with controls. RNA was isolated 24 hours later from peripheral blood mononuclear cells. Gene expression patterns in blood were then sur- veyed, using triplicate Affymetrix oligonucleotide ar- rays. The results are consistent with the suggestion that different disease states produce distinct genomic re- sponses in peripheral white blood cells. We propose that the blood genomic response will serve as a finger- print for these and many other medical and neurolog- ical diseases. Materials and Methods Animal Models Adult male Sprague-Dawley rats (Harlan, Indianapolis, IN), weighing 250 to 300gm, were used. Animals were acclimated to the animal quarters at least 3 days before study. Six rats were used for each group: 6 for brain ischemic stroke, 6 for brain hemorrhagic stroke, 6 for kainate-induced seizures, 6 From the 1 Department of Neurology and Neuroscience Program; and 2 Division of Molecular Developmental Biology and Informat- ics, Children’s Hospital Research Foundation, University of Cincin- nati, Cincinnati, OH. Received Jul 9, 2001, and in revised form Aug 29, 2001. Accepted for publication Aug 29, 2001. Published online Nov 1, 2001; DOI/10.1002/ana.10042 Address correspondence to Dr Sharp, Department of Neurology and Neuroscience Program, University of Cincinnati, Vontz Center for Molecular Studies, Room 2327, 3125 Eden Avenue, Cincinnati, OH 45267-0536. E-mail: [email protected] ORIGINAL ARTICLES © 2001 Wiley-Liss, Inc. 699
Transcript

Blood Genomic Responses Differ AfterStroke, Seizures, Hypoglycemia, and

Hypoxia: Blood Genomic Fingerprintsof Disease

Yang Tang, MD,1 Aigang Lu, MD,1 Bruce J. Aronow, PhD,2 and Frank R. Sharp, MD2

Using microarray technology, we investigated whether the gene expression profile in white blood cells could be used asa fingerprint of different disease states. Adult rats were subjected to ischemic strokes, hemorrhagic strokes, sham sur-geries, kainate-induced seizures, hypoxia, or insulin-induced hypoglycemia, and compared with controls. The whiteblood cell RNA expression patterns were assessed 24 hours later using oligonucleotide microarrays. Results showed thatmany genes were upregulated or downregulated at least twofold in white blood cells after each experimental condition.Blood genomic response patterns were different for each condition. These results demonstrate the potential of blood geneexpression profiling for diagnostic, mechanistic, and therapeutic assessment of a wide variety of disease states.

Ann Neurol 2001;50:699–707

With the rapid advance of DNA microarray technol-ogy and sequencing of rodent and human genomesnearing completion, it is possible to examine the ex-pression of thousands of genes in a single sample ofRNA derived from various tissues or organisms.1,2 Mi-croarrays have been used to describe genomic responsesto a variety of different stimuli, including serum stim-ulation,3 iron,4 glucose deprivation,5 and hypoxia.6

Disease-related gene expression patterns are beingdescribed, including genes specific for inflammatorydisease,7 evidence of toxic exposure,8 and multiple scle-rosis.9 Gene expression patterns in tumor tissue are dif-ferent from those in the host,10 as well as different pat-terns in different tumor types.11,12 Genomic responsesin tumor cells from patients with acute myeloid leuke-mia differ from responses in acute lymphoblastic leu-kemia.13 Of note has been the finding that even nor-mal peripheral white blood cells demonstrate differentgene expression patterns when stimulated with differ-ent agents.14,15

These latter findings prompted us to wonderwhether the gene expression patterns in white bloodcells would be distinct in different disease states, andwhether the blood gene expression profile could serveas a fingerprint for various disease states. We chose toexamine the blood gene expression profile because

blood is the most readily accessible tissue in patients.Blood gene expression profiles might make it possibleto diagnose many medical and toxic conditions that arecurrently difficult or impossible to assess. To begin toaddress this, we subjected adult rats to various brain-specific or systemic stimuli, including ischemic strokes,hemorrhagic strokes, sham surgeries, kainate-inducedseizures, hypoxia, and insulin-induced hypoglycemia,and compared them with controls. RNA was isolated24 hours later from peripheral blood mononuclearcells. Gene expression patterns in blood were then sur-veyed, using triplicate Affymetrix oligonucleotide ar-rays. The results are consistent with the suggestion thatdifferent disease states produce distinct genomic re-sponses in peripheral white blood cells. We proposethat the blood genomic response will serve as a finger-print for these and many other medical and neurolog-ical diseases.

Materials and MethodsAnimal ModelsAdult male Sprague-Dawley rats (Harlan, Indianapolis, IN),weighing 250 to 300gm, were used. Animals were acclimatedto the animal quarters at least 3 days before study. Six ratswere used for each group: 6 for brain ischemic stroke, 6 forbrain hemorrhagic stroke, 6 for kainate-induced seizures, 6

From the 1Department of Neurology and Neuroscience Program;and 2Division of Molecular Developmental Biology and Informat-ics, Children’s Hospital Research Foundation, University of Cincin-nati, Cincinnati, OH.

Received Jul 9, 2001, and in revised form Aug 29, 2001. Acceptedfor publication Aug 29, 2001.

Published online Nov 1, 2001; DOI/10.1002/ana.10042

Address correspondence to Dr Sharp, Department of Neurology andNeuroscience Program, University of Cincinnati, Vontz Center forMolecular Studies, Room 2327, 3125 Eden Avenue, Cincinnati,OH 45267-0536. E-mail: [email protected]

ORIGINAL ARTICLES

© 2001 Wiley-Liss, Inc. 699

for insulin–glucose, 6 for hypoxia, 6 untouched controls, and6 sham operations. The blood from 2 rats in each group wascombined into a single sample, so that enough RNA wasavailable for a single microarray. Brain samples from eachanimal in each group were taken from the same region ofparietal neocortex. Therefore, three separate blood RNAsamples (from 6 rats) were placed on three microarrays andthree separate brain RNA samples (from 3 rats) on three mi-croarrays for each group. We used triplicate microarrays be-cause of greater reliability with triplicates.16 We did not per-form replicates of the same samples of RNA because of themanufacturer’s (Affymetrix, Santa Clara, CA) data demon-strating good reliability of the chips when analyses are per-formed on separate chips with the same sample of RNA.

Brain IschemiaTo produce brain ischemia, rats were anesthetized withisoflurane (3% in 21% oxygen and 76% nitrogen), neck skinand muscle were incised, and the left common carotid arterywas isolated. Body temperature was maintained at 37.0 60.2°C with a rectal thermistor connected to a feedbackcontroller-driven heating pad. A 3-0 monofilament nylon su-ture was threaded through the external carotid artery stumpinto the internal carotid artery and up to the stem of themiddle cerebral artery (MCA). The suture was anchored inplace with 4-0 silk to produce a permanent MCA occlusion.Muscle and skin were sutured; once the animals recovered,they were returned to their home cages with food and wateravailable ad libitum. This “suture” model of MCA occlusionproduces reliable infarction in the distribution of theMCA.17,18

Intracerebral HemorrhageTo produce brain hemorrhage, adult rats were anesthetizedwith isoflurane. The scalp was incised and a burr hole drilled0.5mm anterior and 4mm lateral to the bregma. A 25 gaugeneedle was used to deliver 50ml of lysed arterial blood 4mmdeep into the right striatum in 6 animals. The wound wascleaned and sutured. Animals recovered in their home cageswith food and water ad libitum for 24 hours. This type ofbrain hemorrhage results in cell death around the margins ofthe hemorrhage.19

Sham OperationsThese animals were anesthetized and neck and muscle inci-sions performed. The common carotid was isolated, but nosuture was inserted. The wounds were sutured. Once animalsrecovered from anesthesia, they were returned to their homecages with food and water available ad libitum.

Kainate SeizuresRats were injected subcutaneously with 10mg/kg of kainicacid dissolved in 0.9% sterile saline (Sigma, St Louis, MO).Animals that had severe, prolonged generalized seizures werestudied.20 Animals remained in their home cages for 24hours with food and water available ad libitum. Prolongedseizures resulted in injury to neurons in cortex, hippocam-pus, entorhinal cortex, and other brain regions.20,21

Insulin and GlucoseAdult rats were injected with 10U/kg regular insulin subcu-taneously. One-half of the subjects became obtunded, andthe other half became obtunded and had seizures. Approxi-mately 4 to 6 hours after receiving insulin, animals weregiven 20ml of 20% sterile glucose intraperitoneally, repeatedevery hour for 2 hours. Of the 6 animals injected, 4 recov-ered and appeared to be normal. The other 2 animals con-tinued to have intermittent seizures. All animals were sacri-ficed at 24 hours after the insulin injections.

HypoxiaAdult rats were placed in a large plexiglass chamber (BioIn-struments, Redfield, NY) through which 8% oxygen was cir-culated. The oxygen and carbon dioxide concentrations weremonitored continuously. After 6 hours of hypoxia, animalswere returned to normoxia in their home cages for 18 hours.This degree and duration of hypoxia induces the hypoxia-inducible factor (HIF-1) in brain.22

Untouched ControlsRats that had not been handled in any way were used ascontrols. These animals were allowed access to food and wa-ter ad libitum and were exposed to a 12 hour light/12 hourdark cycle, as was the case for the all the other animals inthis study. All experimental and control animals were housedin the same room both before and during the study.

RNA Isolation from Blood and BrainAt 1 day (24 hours) after brain ischemia, intracerebral hem-orrhage, sham surgery, kainic acid injection, hypoxia, or in-sulin–glucose injection, or assignment as an untouched con-trol, all subjects were anesthetized with ketamine (100mg/kg)and xylazine (20mg/kg). Once the animal was deeply anes-thetized, a 20 gauge needle was used to withdraw $5ml ofwhole (heparinized) blood from the left ventricle of theheart. Immediately after this the animal was decapitated andthe brain rapidly removed. The brain was cut coronally atthe level of the bregma, and the motor and sensory cortexdissected up to the cingulate dorsally and the whisker sensorycortex ventrally.

Immediately after withdrawal of blood, the white cellswere isolated from whole blood using Ficoll-Paque Plus(Amersham, Piscataway, NJ). The Ficoll method has beenused in previous microarray studies13 and is used to isolaterelatively pure populations of mononuclear cells. Total RNAfrom both the parietal cortex and the white cells was isolatedwith TRIZOL Reagent (Human Genome Sciences, Rock-ville, MD) and purified with RNeasy mini kit (Qiagen, Va-lencia, CA). RNA samples were quantified by spectropho-tometry and stored at 280°C for microarray and reversetranscription–polymerase chain reaction (RT-PCR) studies.

Microarray AnalysisSample labeling, hybridization to arrays, and image scanningwere carried out according to the manufacturer’s instructions(Affymetrix). The rat U34A array used contained more than7,000 genes and 1,000 ESTs (expressed sequence tags, re-ferred to as genes in the following text). Each gene on thearray was assessed using 16 probe pairs. Each probe pair con-

700 Annals of Neurology Vol 50 No 6 December 2001

sisted of an oligomer (25 bases long) perfectly complemen-tary to a particular message (called the perfect match [PM])and a companion oligomer identical to the PM probe exceptfor a single base difference in a central position (called themismatch [MM] probe). The mismatch probe served as acontrol for hybridization specificity and helped subtract non-specific hybridization. After hybridization intensity data werecaptured, the Affymetrix Genechip software calculated inten-sity values for each probe cell and used these probe cell in-tensities to calculate an average intensity for each gene (av-erage difference in intensities between PM and MM cells; theaverage intensity directly correlates with mRNA abundance).The software also gave each gene a qualitative assessment of“present” or “absent” based on a voting scheme, with thenumber of instances in which the PM signal is significantlylarger than the MM signal across the whole probe set. Onlythose genes scored as present on three of three chips for eachexperimental condition were used for the analysis.

Before comparing any two measurements, a scaling proce-dure was performed so that all signal intensities on an arraywere multiplied by a factor that makes the average intensityvalue for each array equal to a preset value of 1,500. Scalingcorrects for any interarray differences or small differences insample concentration, labeling efficiency, or fluorescence de-tection and makes interarray comparisons possible.

The ischemic stroke, hemorrhagic stroke, sham surgery,hypoxia, kainate, and insulin–glucose results were all com-pared with the untouched controls. First, the numbers of

genes that had a twofold increase or more (upregulated,greater than twofold) or a twofold decrease or more (down-regulated, greater than twofold) average expression in the ex-perimental compared with control samples were determined(Table, .twofold). The upregulated and downregulatedgenes for each of the experimental conditions all had three“present” calls, and their average difference had to be morethan 100 on all three experimental chips in the group. Be-cause there is no agreed-upon detection threshold for thechips, we have also listed genes that had three “present” callsand that had an average difference value of more than 1,000(see Table). This was done to ensure that the results were notsignificantly affected or skewed by the lower value chosen forthe analysis of fold changes. To provide additional indices ofthe reliability of the differences between the experimentalvalues and untouched control values, the “greater than two-fold” list of genes was used to derive two additional lists (seeTable). In the .twofoldb list, the numbers of genes areshown when all of three measurements in a specific experi-mental condition were greater than those in the correspond-ing control condition, or likewise when all three were lessthan in controls. Gene numbers for which a statistically sig-nificant difference (p , 0.05, Student’s t test) was estab-lished between experimental and control conditions are listedin the last column of the table.

A cluster analysis (Genespring software; Silicon Genetics,Redwood City, CA) was performed on all the genes that (1)were upregulated or downregulated more than twofold in the

Table. Numbers of Regulated Genes for Each Condition Compared to Controls

Condition

No. ofGenes

Addressed

No. of Regulated Genes

.Twofolda .Twofoldb .Twofoldc

Raw.100

Raw.1,000

Raw.100

Raw.1,000

Raw.100

Raw.1,000

Brain ischemia vs untouch Upregulated 2,507 25 20 15 12 8 8Downregulated 2,538 98 73 33 24 10 9

Brain hemorrhage vsuntouch

Upregulated 2,109 27 23 10 6 4 3

Downregulated 2,538 193 121 89 63 23 17Kainate-induced seizure vs

untouchUpregulated 1,681 106 97 63 55 16 13

Downregulated 2,538 311 214 195 139 27 22Insulin–glucose vs

untouchUpregulated 1,894 94 86 43 40 25 25

Downregulated 2,538 231 161 165 116 52 45Hypoxia vs untouch Upregulated 1,761 139 134 78 75 19 18

Downregulated 2,538 294 214 145 99 12 9Sham vs untouch Upregulated 2,408 40 34 26 23 18 18

Downregulated 2,538 162 105 82 64 50 43

Numbers of white blood cell genes that were present and regulated 24 hours after brain ischemia, brain hemorrhage, kainate-induced seizures,insulin–glucose, or hypoxia as compared with untouched control animals. The number of genes addressed indicates genes and ESTs that havethree “present” calls in three chips for that group. Upregulated genes have three “present” calls in experimental chips; downregulated genes havethree “present” calls in control chips (ie, 2,538 for every downregulated group). Genes are listed based on whether the raw average valuemeasured on the chips was either .100 or .1,000.aGenes up- or downregulated more than twofold.bGenes regulated more than twofold and where every value in the experimental group was either greater or every value was less than any of thethree measurements in the control group.cGenes regulated more than twofold. No overlap between the 3 experimental data and 3 control data points; Student’s t test showed p ,0.05for comparison of experimental with control values.

Tang et al: Blood Genomic Fingerprints of Disease 701

experimental conditions compared with the untouched con-trol condition; and (2) showed no overlap of the values inthe experimental condition versus the values in the un-touched controls (see Table). Separate cluster analyses wereperformed for the 605 genes for which the raw values weregreater than 100; and for the 474 genes in which the rawvalues were greater than 1,000 (see Table, .twofold). Hier-archical clustering was performed, and a standard correlationcoefficient of 0.95 was used as the measure for significantstatistical similarity. Genes having similar expression patternsacross the 7 groups were clustered together. The branchingbehavior of the tree was controlled using a separation ratiosetting of 0.5 and a minimum distance setting of 0.001.

Quantitative Reverse Transcriptase-PolymeraseChain ReactionReal-time RT-PCR was performed on five selected genes us-ing the 5700 Sequence Detection System (Applied Biosys-tems, Foster City, CA). All primers and probes were de-signed using Primer Express 1.0. One-step RT-PCR wasperformed according to manufacturer’s instructions (Taqmangold RT-PCR Kit; Applied Biosystems).

ResultsBlood Genomic ResponseOf the 8,740 genes surveyed, 19% (1,681 of 8,740 inthe kainate group) to 29% (2,538 of 8,740 in the un-touched group) are “present” in mononuclear whiteblood cells (see Table). Of the “present” transcripts ineach group, some were induced or suppressed at leasttwofold by different types of systemic and organ-specific stimuli or injury. The Table shows the numberof genes that were upregulated or downregulated ineach condition based on whether the raw measurementwas greater than 100 or greater than 1,000. For exam-ple, if a twofold change criteria is used for genes with araw value of more than 100, as few as 25 of 2,507transcripts are upregulated by brain ischemia in whiteblood cells, and as many as 311 of 2,538 transcripts aredownregulated by kainate-induced seizures (see Table).If more stringent criteria are used, as few as 6 of 2,109(0.3%) transcripts are upregulated by brain hemor-rhage and as many as 139 of 2,538 transcripts (5.5%)downregulated by kainate-induced seizures (see Table).No matter how stringent the criteria, including statis-tical analysis of genes expressed with raw values greaterthan 1,000, significant numbers of transcripts were reg-ulated by each condition.

Genes Regulated by Multiple ConditionsMany genes upregulated or downregulated by each ex-perimental condition were modulated in two or moreof the groups. For example, of 25 transcripts upregu-lated twofold by brain ischemia, 6 were also inducedby brain hemorrhage and 4 by brain ischemia, brainhemorrhage, and sham surgery (Fig 1a). Because ani-mals subjected to ischemic and hemorrhagic stroke andsham surgeries all had surgery while anesthetized, someupregulated (see Fig 1a) and downregulated (see Table)genes in these groups may be due to the anesthesia orthe surgery. Similarly, of 311 genes downregulated atleast twofold by kainate (see Table and Fig 1b), 102are downregulated by insulin–glucose and 62 by kai-nate, insulin–glucose, and hypoxia. The 40 genes thatare regulated by kainate and insulin–glucose, but notby hypoxia, may be accounted for by the fact that thekainate animals and several of the insulin–glucose ani-mals had seizures. Similar mechanisms may mediatethe induction of these genes when they are regulated in2 or more groups. Some transcripts were induced or

Fig 1. Venn diagrams show the numbers of genes upregulated(a) or downregulated (b) in several groups. (a) Numbers ofgenes that were upregulated more than twofold in blood 24hours after brain ischemia (BI), brain hemorrhage (BH), andsham surgery (S), compared with untouched controls. (b)Numbers of genes that were downregulated more than twofoldin blood 24 hours after kainate (K), insulin–glucose (IG),and hypoxia (H), compared with untouched controls.

702 Annals of Neurology Vol 50 No 6 December 2001

suppressed by all the experimental conditions whencompared with controls. For example, porphobilinogendeaminase (Accession no. X06827) was induced inblood white cells after all the stimuli mentioned above(Fig 2). It is possible that systemic stress and cat-echolamines induce a number of common genes foreach condition (Fig 3).

Genes Specifically Regulated by Each ConditionGenes that were upregulated or downregulated in onlyone condition were then identified to determinewhether there is a specific blood genomic response foreach experimental condition. For most experimentalconditions examined here, there was a group of specif-

ically regulated genes. For example, though there wasonly one gene specifically upregulated by brain hemor-rhage (Fig 4a), there were over a dozen genes specifi-

Fig 2. Microarray results compared with real-time reversetranscription-polymerase (RT-PCR) results. Adult rats weresubjected to sham operations (S); brain ischemia (BI) pro-duced by a permanent middle cerebral artery occlusion; brainhemorrhage (BH) produced by intracerebral infusions of lysedblood; kainate (K)-induced seizures; hypoxia (H) produced byexposure to 8% oxygen for 6 hours; and insulin-induced hypo-glycemia followed by glucose administration (IG). One daylater (24 hours), total RNA from the blood of three pairs ofanimals for each condition was isolated. The RNA expressionlevels of five selected genes were assessed by both microarraysand real-time RT-PCR and normalized to that of the controls.The fold change (y axis) was calculated as the expression ofthe gene in each condition versus the controls, so that the foldchange for control in every case is 1. (A) X06827 prophobi-linogen deaminase; (B) M60666 a-tropomyosin; (C) U39875EF-hand calcium binding protein; (D) AF045464 aflatoxinB1 aldehyde reductase; (E) L00603 monoamine transporter.

Fig 3. Hierarchical clustering of genes with raw values of.100 (A; n 5 605) or .1,000 (B; n 5 474) that were reg-ulated at least twofold after sham surgery, brain ischemia, brainhemorrhage, kainate-induced seizures, hypoxia, and insulin–glucose, compared with untouched controls. The clustered geneswere derived from those in the Table (.2-fold data), andhence showed no overlap between experimental and untouchedcontrol values. Genes that show similar expression patterns acrossdifferent treatments cluster together. Red 5 upregulation; yellowgreen 5 little change; deep green 5 downregulation.

Tang et al: Blood Genomic Fingerprints of Disease 703

Figure 4

704 Annals of Neurology Vol 50 No 6 December 2001

cally downregulated by brain hemorrhage comparedwith the other groups (see Fig 4b). Similarly, therewere specifically upregulated and downregulated tran-scripts for kainate-induced seizures (see Fig 4c and d),hypoxia (see Fig 4e and f), and insulin–glucose treat-ment (see Fig 4g and h), as compared with the otherexperimental groups. However, there was no differen-tially expressed gene specific for brain ischemia (see Fig4i and j).

Unique Gene Expression Profile in PeripheralWhite CellsBecause there may not be specifically regulated genesfor many disease conditions, we looked for a uniquepattern of gene expression for each of the experimentalconditions, including brain ischemia. For example, Fig-ure 4 shows gene profiles that could be used to distin-guish gene expression between sham surgery, brainischemia, and brain hemorrhage (see Fig 4i and j). Abroader view of global expression profiling was accom-plished using cluster analysis. This approach showedthat each experimental condition produced a uniquegene expression pattern in white cells (see Fig 3). Toensure that our conclusions were not affected by thethreshold of gene expression chosen, separate clusteranalyses were performed for the genes that had raw val-ues of greater than 100 or greater than 1,000 and wereregulated more than twofold without any overlap be-tween experimental and control values (see Table). Forboth analyses, the overall clustering pattern was similar.Some gene clusters were regulated by only one treat-ment, while others were regulated by two or moretreatments. For the seven global expression profilesshown, no profile was identical. This was true whetherthe raw value threshold was greater than 100 or greaterthan 1,000. The different gene expression patterns, in-cluding sham versus untouched controls, emphasize theuniqueness of the different blood genomic responsesfor each condition. The overall similarities in the bloodgenomic response patterns in each of the six experi-mental conditions compared with the untouched con-trols suggest that similar genes are induced in responseto stress in all of these conditions.

Examples of the upregulated genes point to theunique pattern of gene expression in each condition.

Some of the genes upregulated greater than twofold inblood mononuclear cells after brain ischemia comparedwith untouched controls included transferrin receptor(M58040), protease 1 (S69206), pituitary transform-ing gene (U73030), vesicular monoamine transporter(L00603), junD (D26307), patched (AF079162), andthromboxane A2 receptor (D32080). Transcripts up-regulated greater than twofold in the blood after brainhemorrhage included transferrin receptor (M58040),rat-related rab 1B protein (X13905), replication factor C(AF030050), and vascular protein tyrosine phosphatase-1rDEP-1 (U40790). Genes upregulated by insulin–glucose in blood included glucose-regulated 94 relatedprotein (AB003515), NaPi-2 b (AB013454), dihydro-lipoamide acetyltransferase (D10655), 12-lipoxygenase(L06040), asparagine synthetase (U07201), carbonic an-hydrase II (U60578), ADP-ribosylation factor (L12384),syntaxin 5 (L20822), and RNA polymerase II transcrip-tion factor SIII p18 subunit (L42855). Transcripts up-regulated more than twofold by hypoxia in blood in-cluded ATP synthase subunit (D13120), mitochondriallong-chain enoyl-CoA hydratase (D16478), mitochon-drial long-chain 3-ketoacyl-CoA thiolase (D16479),NADH-ubiquinone oxidoreductase (D86215), long-chain acyl-CoA synthetase (D90109), phospholipidhydroperoxide glutathione peroxidase (L24896), a-tropomyosin (M15474), proton pump polypeptide(M58758), bcl-xs (S78284), asparagine synthetase(U07201), EF-hand calcium binding protein p22(U39875), and a putative chloride channel (Z36944).Transcripts induced more than twofold after kainateinduced seizures included PSD-Zip45 (AB017140),muscarinic receptor m2 (AB017655), putative phero-mone receptor (AF016178), neuron glucose transporterGLUT3 (D13962), NMDA receptor glutamate-bindingsubunit (S61973), AMP-regulated phosphoprotein(S65091), and dopamine D3 receptor (X53944). Anumber of transcripts, including the 50 kDa glyco-protein (RH50; AB015194), were upregulated by sev-eral conditions. Although space precludes the listingof downregulated transcripts, these may be as impor-tant as those that are upregulated. For example,poly(ADP-ribose)polymerase (U94340), notch (X57405),tricarboxylate carrier (S70011), a-fodrin (AF084186),retinoblastoma protein (D25233), p38 mitogen acti-

Š Fig 4. Selected genes that were specifically upregulated (a, c, e, g, i) or downregulated (b, d, f, h, j) greater than twofold in bloodafter various experimental conditions, compared with controls (C). Adult rats were subjected to sham operations, brain ischemiaproduced by a permanent middle cerebral artery occlusion, brain hemorrhage produced by intracerebral infusions of lysed blood,kainate-induced seizures, hypoxia produced by exposure to 8% oxygen for 6 hours, and hypoglycemia produced by insulin injectionsthat was reversed by glucose (x-axis labels as in Fig 2). Twenty four hours later, total RNA from the blood of three pairs of ani-mals for each condition was isolated and gene expression assessed using Affymetrix rat oligonucleotide microarrays. Each of the panelsshows the fold changes of gene expression for one or more genes that were upregulated or downregulated as compared with controlsafter brain hemorrhage (a and b), kainic acid (c and d), hypoxia (e and f), insulin–glucose (g and h), and sham operations, brainischemia, and brain hemorrhage (I and j). Error bars demonstrate the range of values measured in each of the 3 subjects.

Tang et al: Blood Genomic Fingerprints of Disease 705

vated protein kinase (U91847), b-adrenergic receptorkinase-1 (M87854), and others are downregulated inmononuclear white blood cells 24 hours after 6 hoursof 8% hypoxia. Lists of all the upregulated and down-regulated genes derived from the Table are provided assupplementary information at http://www.interscience.wiley.com/jpages/0364-5134/suppmat/index/html.

Blood Genomic Markers for Brain InjuryOne of the original hypotheses that stimulated thesestudies was that cell death in the brain might modulatespecific transcripts in white blood cells that would bedifferent from stimuli that do not lead to cell death.Two transcripts were induced after brain ischemiaand kainate, and variably in insulin–glucose and brainhemorrhage: vascular protein tyrosine phosphatase-1rDEP-1 (U40790) and vesicular monoamine trans-porter (L00603). In addition, the histamine H2receptor (S57565) was downregulated after brain isch-emia, brain hemorrhage, kainate-induced seizures, andinsulin–glucose.

Correlation of Microarray Data With PolymeraseChain Reaction ResultsQuantitative RT-PCR was performed on five selectedgenes (see Fig 2). The RT-PCR results showed excel-lent agreement with the corresponding microarray re-sults (porphobilingen deaminase; a-tropomyosin; EF-hand calcium binding protein; and the monoaminetransporter). For one gene (aflatoxin B1 aldehyde re-ductase), RT-PCR results showed a much greaterchange as compared with the microarray results (seeFig 2D). However, both the microarray and the RT-PCR results for aflatoxin B1 aldehyde reductaseshowed the same pattern of expression in each of theconditions: low in control; somewhat higher in sham;little change in brain ischemia, brain hemorrhage, andhypoxia; and the highest levels of expression in kainateand insulin–glucose (see Fig 2).

DiscussionThe major finding of this study is that different pat-terns of gene expression occur in the peripheral whiteblood cells 1 day after ischemic stroke, hemorrhagicstroke, seizures, hypoxia, and hypoglycemia, as com-pared with control animals. The results imply that dif-ferences of gene expression in peripheral white bloodcells could be used in the clinic to diagnose the occur-rence of these events 1 day (or possibly many days)previously. It might be possible to diagnose recent hy-poglycemia in diabetic patients, recent seizures in pa-tients with epilepsy, and recent hypoxia in patientswith syncope or cardiovascular disease. Because of thelarge numbers of genes expressed by white blood cells,and differences that would be manifested over time, itis possible that there is a genomic fingerprint for a

wide variety of diseases and toxic states that could beused to diagnose and monitor these conditions.

It is not too surprising that systemic hypoxia andinsulin produce changes of gene expression in whiteblood cells, since most cells in the body including thewhite cells would have been hypoxic or hypoglycemic.It is also not surprising that there would be genes spe-cifically induced by hypoxia and insulin–glucose, asthere are both oxygen or hypoxia-inducible genes23 andglucose-regulated proteins.24 The greatest surprise wasthe upregulation and downregulation of genes withorgan-specific injury. That is, genes are induced orsuppressed in white blood cells after brain ischemia,brain hemorrhage, and kainate-induced seizures. Themechanisms by which different brain injuries might af-fect specific gene expression in white blood cells maybe related to immune surveillance. However, the spe-cific factors that would lead to the downregulation ofso many genes in each of the conditions remain un-clear. There appears to be considerable specificity, asischemic and hemorrhagic stroke and kainate-inducedseizures produced different patterns of gene downregu-lation in the blood. This could relate to inhibition ofspecific signaling pathways. Alternatively, the patternsof altered gene expression could reflect altered functionor apoptosis of certain classes of white blood cells. Asno attempt was made to examine gene expression pat-terns in different classes of white blood cells, it is pos-sible that the gene expression changes reflect differentclasses of white blood cells that are affected differen-tially by each condition.

Alternatively, the differences of mononuclear cellgene expression after ischemic stroke, hemorrhage, sei-zures, and hypoglycemia could relate to unique im-mune responses to injury. There is likely a unique im-mune response to dying neurons, glia, and vasculatureafter ischemic infarction; to blood elements after hem-orrhage; and possibly to the death of different selec-tively vulnerable neurons produced by seizures and hy-poglycemia.

Although only a limited number of disease condi-tions were studied, the data show that the bloodgenomic response was unique for each. The best exam-ple of this difference was the unique blood genomicresponse observed in sham-operated control animalsthat differed from untouched controls. Even thoughthere are genes that are specifically regulated by each ofthe conditions, we propose that it is the compositeblood genome response—all the upregulated anddownregulated genes—that will provide a genomic fin-gerprint that can be distinguished for many differentdisease-related states.

These studies add to a growing literature demon-strating the use of microarray-based gene expressionpatterning in the molecular classification of dis-ease.13,14 However, to our knowledge, this is the first

706 Annals of Neurology Vol 50 No 6 December 2001

effort to show the potential of using blood genomicresponses as markers for end-organ diseases. Extendingthese studies to humans is the next step. However, thediverse genetic background and multiple environmen-tal factors that affect humans could result in more vari-ation than we have seen in rats, which could make pat-tern recognition of blood genomic responses to diseasein humans more difficult.

These studies were supported by the National Institutes of Health(NS28167, NS38084, and NS38743) and by the Bugher Founda-tion (0070006N) from the American Heart Association (to F.R.S.).

References1. Brown PO, Botstein D. Exploring the new world of the ge-

nome with DNA microarrays. Nature Genet 1999;21:33–37.2. Lipshutz RJ, Fodor SP, Gingeras TR, et al. High density syn-

thetic oligonucleotide arrays. Nature Genet 1999;21:20–24.3. Iyer VR, Eisen MB, Ross DT, et al. The transcriptional pro-

gram in the response of human fibroblasts to serum. Science1999;283:83–87.

4. Yun CW, Ferea T, Rashford J, et al. Desferrioxamine-mediatediron uptake in Saccharomyces cerevisiae. Evidence for two path-ways of iron uptake. J Biol Chem 2000;275:10709–10715.

5. Ferea TL, Botstein D, Brown PO, et al. Systematic changes ingene expression patterns following adaptive evolution in yeast.Proc Natl Acad Sci U S A 1999;96:9721–9726.

6. Ter Linde JJ, Liang H, Davis RW, et al. Genome-wide tran-scriptional analysis of aerobic and anaerobic chemostat culturesof Saccharomyces cerevisiae. J Bacteriol 1999;181:7409–7413.

7. Heller RA, Schena M, Chai A, et al. Discovery and analysis ofinflammatory disease-related genes using cDNA microarrays.Proc Natl Acad Sci U S A 1997;94:2150–2155.

8. Pennie WD. Use of cDNA microarrays to probe and under-stand the toxicological consequences of altered gene expression.Toxicol Lett 2000;112/113:473–477.

9. Whitney LW, Becker KG, Tresser NJ, et al. Analysis of geneexpression in mutiple sclerosis lesions using cDNA microarrays.Ann Neurol 1999;46:425–428.

10. Perou CM, Jeffrey SS, van de Rijn M, et al. Distinctive geneexpression patterns in human mammary epithelial cells andbreast cancers. Proc Natl Acad Sci U S A 1999;96:9212–9217.

11. Ross DT, Scherf U, Eisen MB, et al. Systematic variation ingene expression patterns in human cancer cell lines. NatureGenet 2000;24:227–235.

12. Scherf U, Ross DT, Waltham M, et al. A gene expression da-tabase for the molecular pharmacology of cancer. Nature Genet2000;24:236–244.

13. Golub TR, Slonim DK, Tamayo P, et al. Molecular classifica-tion of cancer: class discovery and class prediction by gene ex-pression monitoring. Science 1999;286:531–537.

14. Alizadeh AA, Eisen MB, Davis RE, et al. Distinct types of dif-fuse large B-cell lymphoma identified by gene expression pro-filing. Nature 2000;403:503–511.

15. Staudt LM, Brown PO. Genomic views of the immune system.Annu Rev Immunol 2000;18:829–859.

16. Lee ML, Kuo FC, Whitmore GA, et al. Importance of replica-tion in microarray gene expression studies: statistical methodsand evidence from repetitive cDNA hybridizations. Proc NatlAcad Sci U S A 2000;97:9834–9839.

17. Zarow GJ, Karibe H, States BA, et al. Endovascular suture oc-clusion of the middle cerebral artery in rats: effect of sutureinsertion distance on cerebral blood flow, infarct distributionand infarct volume. Neurol Res 1997;19:409–416.

18. Rajdev S, Hara K, Kokubo Y, et al. Mice overexpressing ratheat shock protein 70 are protected against cerebral infarction.Ann Neurol 2000;47:782–791.

19. Matsushita K, Meng W, Wang X, et al. Evidence for apoptosisafter intercerebral hemorrhage in rat striatum. J Cereb BloodFlow Metab 2000;20:396–404.

20. Zhang X, Gelowitz DL, Lai CT, et al. Gradation of kainic acid-induced rat limbic seizures and expression of hippocampal heatshock protein-70. Eur J Neurosci 1997;9:760–769.

21. Zhang X, Boulton AA, Yu PH. Expression of heat shockprotein-70 and limbic seizure-induced neuronal death in the ratbrain. Eur J Neurosci 1996;8:1432–1440.

22. Bergeron M, Yu AY, Solway KE, et al. Induction of hypoxia-inducible factor-1 (HIF-1) and its target genes following focalischaemia in rat brain. Eur J Neurosci 1999;11:4159–4170.

23. Semenza GL. Perspectives on oxygen sensing. Cell 1999;98:281–284.

24. Massa SM, Swanson RA, Sharp FR. The stress gene response inbrain. Cerebrovasc Brain Metab Rev 1996;8:95–158

Tang et al: Blood Genomic Fingerprints of Disease 707


Recommended