Gene expression using microarrays intransplant recipients at risk of EBVlymphoproliferation after organtransplantation: Preliminaryproof-of-concept

PTLD is a potentially devastating complicationof organ transplantation. This entity is caused bythe EBV, a member of the herpes group ofviruses. This virus has the ability to transformand immortalize B lymphocytes, resulting in thepotential for uncontrolled lymphoproliferation(1–3). In the normal host, such proliferation islimited by EBV-specific CTL activity. However,in transplant patients, immunosuppressive med-ications adversely affect immune function and

Allen U, Barton M, Beyene J, Hu P, Khodai-Booran N, Hebert D,Dipchand A, Ng V, Solomon M, Grant D, Fecteau A, Ngan B, Read S,Zielenska M, Weitzman S. Gene expression using microarrays intransplant recipients at risk of EBV lymphoproliferation after organtransplantation: Preliminary proof-of-concept.PediatrTransplantation2009:13:990–998.�2009JohnWiley&SonsA/S.

Abstract: We hypothesized that aspects of the virus–host interactioncould be measured to help predict progression to EBV–PTLD. Weexamined the spectrum of host genes differentially expressed and anyrelevant clustering in children at risk of EBV lymphoproliferation afterorgan transplantation. We compared the genes expressed among pa-tients with different levels of viral loads. Gene expression was measuredby microarray analysis of RNA from CD19+ B lymphocytes using theHuman Genome U133 Plus 2.0 GeneChip. Among 27 samples from 26transplant recipients, the viral load categories were: low or undetectableloads (LU), n = 14; high or intermediate loads (HI), n = 13. Therewere seven healthy EBV-seropositive (P) and -seronegative controls (N).Median time of post-transplantation was 0.5 yr (range 0.1–3.8). Weidentified 24–54 differentially expressed genes in each of four compari-sons of HI vs. P, LU vs. P, HI vs. LU, and P vs. N. We identifiedpatterns of 563 gene expressions, creating five clusters aligned withlevels of viral load. PTLD occurred in four of five clusters. In summary,we demonstrated varying degrees of alignment between levels of VL andgene clusters. Analyses for differential expression of genes showed genesthat could be implicated in the pathogenesis of EBV–PTLD.

Upton Allen1,2,3, Michelle Barton1*,Joseph Beyene3*, Pingzhao Hu4*,Nasser Khodai-Booran1, DianeH�bert2,5, Anne Dipchand2,6, VickyNg2,7, Melinda Solomon2,8, DavidGrant2,9, Annie Fecteau2,9, Bo Ngan9,Stanley Read1, Maria Zielenska10 andSheila Weitzman11

1Division of Infectious Diseases, The Department ofPediatrics, 2The SickKids Transplant Centre, 3ChildHealth Evaluative Sciences, 4The Center for AppliedGenomics, Divisions of 5Nephrology, 6Cardiology,7Gastroenterology, Hepatology and Nutrition,8Respiratory Medicine, The Department of Pediatrics,9The Department of General Surgery, 10TheDepartment of Pediatric Laboratory Medicine,11Division of Haematology/Oncology, The Departmentof Pediatrics, Hospital for Sick Children, University ofToronto, Toronto, ON, Canada

Key words: Epstein–Barr virus – lymphoproliferativedisorder – gene expression – microarray

Dr. Upton D. Allen, Division of Infectious Diseases,The Hospital for Sick Children, 555 UniversityAvenue, Toronto, Ontario, Canada M5G 1X8Tel.: +1 416 813 8129Fax: +1 416 813 8404E-mail: upton.allen@sickkids.ca*Joint second author status: Michelle Barton, JosephBeyene, and Pingzhao Hu.

Accepted for publication 18 February 2009

Abbreviations: CCL, chemokine (C-C motif) ligand; CCR5,chemokine (C-C motif) receptor 5; CTL, cytotoxic T lym-phocyte; EBV, Epstein–Barr virus; EIF1AY, eukaryotictranslation initiation factor 1A, Y-linked; FDR, false dis-covery rate; GNLY, granulysin; LMP, latent membraneprotein; LPE, local pooled error; NK, natural killer; PBMC,peripheral blood mononuclear cell; PCR, polymerase chainreaction; PTLD, post-transplant lymphoproliferative dis-order; SLAM, signaling lymphocyte activation molecule;TH1 and 2, T helper cell type 1 and 2.

Pediatr Transplantation 2009: 13: 990–998 � 2009 John Wiley & Sons A/S.

Pediatric TransplantationDOI: 10.1111/j.1399-3046.2009.01193.x

990

CTL activity and create the potential for uncon-trolled lymphoproliferation.Pediatric patients are more likely to develop

PTLD because they often develop primary EBVinfection after transplantation. Thus, they gointo transplantation with no existing CTL activ-ity directed at EBV (4). This is in contrast toadult patients who are often EBV-seropositiveprior to transplantation and therefore entertransplantation with some degree of EBV-direc-ted CTL activity.The outcome of PTLD is highly variable, with

the chances of a successful outcome being depen-dent on early recognition of patients at risk. Tothis end, viral load measurements have been usedto monitor patients at risk of PTLD, given that ithas been shown that elevated loads often precedethe onset of PTLD. However, viral load mea-surements have low specificity as this relates tothe ability to predict which patients will developPTLD (5–7). Studies are needed to improve ourability to predict PTLD, given that elevations ofviral load represent a sensitive but poorly specificindication of the subsequent development ofPTLD. It is desirable to be able to identify thepatients in whom elevation of viral load is likelyto culminate in PTLD, with such identificationbeing done at an early stage prior to the devel-opment of frank PTLD. We questioned whetherthere are aspects of the viral–host interaction thatcould be measured to better indicate if EBVinfection of B lymphocytes will result in lympro-proliferation and the likelihood of PTLD. In thisregard, potential targets to be measured could bevirally encoded genes (e.g., LMP1 and LMP2) orthe genes aligned with the TH1/TH2 profiles. Ourfocus for this study was the latter. A TH2 profilewith less CTL activity to control EBV is favorableto the onset of EBV-related lymphoproliferation.Given the potential benefit of microarray tech-nology in characterizing gene expression in avariety of clinical settings, including hematologicmalignancies, it offers promise in evaluating EBVlymphoproliferation.The study described herein was one of a series of

studies aimed at examining the factors that mightbe predictive of EBV lymphoproliferation afterorgan transplantation. In this particular study,our objectives were threefold. First, to establishthat RNA of adequate quality and quantity formicroarray analyses could be reliably obtainedfrom peripheral blood lymphocytes obtainedfrom pediatric transplant recipients. Second, todetermine the spectrum of genes differentiallyexpressed and any relevant clustering in childrenat risk of EBV lymphoproliferation after organtransplantation. Third, to compare the spectrum

of genes expressed in transplant recipients withdifferent levels of viral loads. The paper ispresented largely as a proof-of-concept study thatwill facilitate further studies in this area.

Methods

Study setting

The study center, the Hospital for Sick Children, Toronto, isa tertiary–quaternary pediatric institution where the cate-gories of transplant patients include renal, cardiac, liver,lung, and small bowel recipients. The project was approvedby the Institutional Research Ethics Board.

Study design and target population

Transplant patients were eligible for this study if they wereEBV donor-seropositive and recipient-seronegative, orrecipient-seropositive regardless of donor status. We pro-spectively enrolled individuals with different categories ofviral loads as follows: undetectable viral loads, low viralloads, intermediate viral loads, and high viral loads. Thesecategories of viral loads were based on our assay as indi-cated below. Gene expression profiles in the transplantpatients were compared with those among control subjectswho were EBV-seropositive and EBV-seronegative. Thelatter two groups consisted of healthy adolescents or adults.

Laboratory procedures

Viral load measurementsEBV loads were measured using a PCR-based semiquanta-tive assay (5). Using this method, the infected cell burden ofPBMCs was assessed. The virus load was measured in therange of <1 infected cell per 106 to >1000 infected cells per106 PBMCs. The categorization of viral loads for this studywas as follows:

Undetected loads: 0 infected cells/106 PBMCsLow load: <100 infected cells/106 PBMCsIntermediate loads: >100 but <1000 infected cells/106

PBMCsHigh viral loads: >1000 infected cells/106 PBMCs

We have previously demonstrated that PTLD is most likelyto be associated with our highest category of viral load (5).

Microarray analyses

RNA extraction: RNA was extracted from CD19+peripheral blood lymphocytes. Briefly, EDTA-containingblood collection tubes were used to collect 10 mL ofperipheral venous blood from each subject. Blood sampleswere then transported to the laboratory for same-dayprocessing. Lymphocytes were separated using Ficoll-Hypaque density centrifugation using standard methodol-ogy. This was followed by a platelet removal procedure.CD19+ B cells were magnetically labeled with CD19MicroBeads. A suspension of cells containing the labeledCD19+ was loaded onto a MACS column and separationachieved by using the magnetic field of a MACS separator(Miltenyi Biotec, Auburn, CA, USA), followed by elutionof the CD19+ from the column. Total RNA was extractedfrom CD19+ cells using the RNeasy� Mini Kit

Gene expression and EBV lymphoproliferation

991

(QIAGEN) according to the manufacturer�s instructions(8). RNA quality and quantity were verified by using abioanalyser (Agilent 2100 Bioanalyser; Agilent Technolo-gies, Inc., Santa Clara, CA, USA).

Microarray analysis: The RNA samples that were extrac-ted from CD19+ B lymphocytes were subjected to micro-array analysis using the Human Genome U133 Plus 2.0GeneChip (Affymetrix, Inc., Santa Clara, CA, USA).Normalization – We used robust multi-array average (9)

methodology to normalize and summarize the data.Gene filtering – The original data had more than 54 000

probe sets. We included only the genes on the microarraywhose expression was highly variable among samples. Wedefined genes that were highly variably expressed withexpression values that were higher by a factor of at least k orlower by a factor of at least 1/k than the average expressionof all samples in at least three of the samples. Based on theabove objectives, for clustering analysis, we set k = 1.4 andthen we obtained 563 genes which met these requirements.For supervised group comparison analysis, we analyzed allthe probe sets. The summarized and normalized data wereimported into the R statistical package (10) and analyzedusing the LPE test (11).To compare the spectrum of genes expressed in transplant

recipients with different levels of viral load, we first cate-gorized patients according to their viral load trends in thethree months preceding the sampling for microarray testing.These patients with high, intermediate, low and undetect-able viral loads were compared among themselves as well aswith healthy controls. For these comparisons, we wereparticularly interested in genes that were differentiallyexpressed and which were aligned with TH1/TH2 cytokinesas well as those known to be associated with hematologicalmalignancies. We computed moderated t-statistics for eachgene based on the LPE test (11), which is based on poolingerrors within genes and between replicate arrays for genes inwhich expression values are similar. The authors demon-strated that the method has advantages to identify subtleexpression changes over other methods. The p-values ob-tained were corrected using Benjamini-Hochberg method(12). The genes were ranked on the basis of the correctedp-values (12). We used the average-linkage hierarchicalclustering method (13) with correlation as the similaritymeasure to identify patterns of the 563 gene expression.

Results

Descriptive characteristics

Thirty-four samples for microarray analyses wereobtained from 33 study subjects. Among 27pediatric samples obtained from 26 organ trans-plant recipients, the corresponding viral loadcategories were as follows: low or undetectableloads, n = 14; high or intermediate loads,n = 13. There were seven healthy control sub-jects. The median age of transplant recipi-ents = 12.1 yr (range 1–16.9); male:femaleratio = 1:1.2. Organ transplanted: kidney(n =13), liver (n = 9), heart (n = 3), and lung(n = 1). Median time post-transplantation0.5 yr (range 0.1–3.8). The samples were CD19+ peripheral blood B lymphocytes, from

which RNA was extracted for the microarrayexperiments. The types of immunosuppressiveagents that the patients were receiving at the timeof microarray testing are also summarized inTable 1, where the most frequently used agentswere tacrolimus, corticosteroids, and mycophen-olate.

RNA quality and quantity

Data from the Agilent 2100 Bioanalyzerindicated that we were able to obtain RNA ofadequate quality and quantity from studysubjects. However, to obtain sufficient RNA formicroarray analyses for some samples, tworounds of amplification were required.

Spectrum of genes expressed and relevant clustering

ClusteringUsing the average-linkage hierarchical clusteringmethod with correlation as the similarity mea-sure, we identified patterns of 563 gene expressionprofiles. The number of clusters can be obtainedbased on the levels of the dendrogram. As Fig. 1shows, we can identify five clusters (shown at thetop of the figure from left to right). The datasetincludes four subjects with five episodes of PTLD.In cluster 1, there was one case of PTLD thatoccurred at 17 months after microarray sampling(liver and abdominal nodal sites). Cluster 3 hadone case occurring at two wk after microarraysampling (pulmonary site). Cluster 4 had oneepisode of PTLD occurring 19 months beforemicroarray sampling (lung, maxillary sinuses,and soft palate). Cluster 5 had two cases ofPTLD (cases 3 and 4) that occurred four monthsbefore and 16 months after microarray sampling,respectively.

Cluster 1: UA23-N, UA2-I, UA18-U, UA11-L,UA17-I, UA19-I (PTLD case 1).

Cluster 2: UA5-U, UA6-U, UA10-U, UA13-U,UA14-U

Cluster 3: UA28-H (PTLD case 2, episode #2),UA29-H, C2-P, NKB1-L, NKB2-L

Table 1. Proportions of pediatric transplant recipients receiving specificimmunosuppressive agents when sampled for microarray testing

Drugs # (%)

Tacrolimus 24/27 (88.5)Corticosteroids 23/27 (85.2)Mycophenolate 14/27 (51.9)Daclizumab/ATG/RATS 5/27 (18.5)Azathiaprine 3/27 (11.1)Cyclosporine 0Sirolimus 0

Allen et al.

992

Cluster 4: UA3-U, UA25-I, UA26-L, NKB8-H(PTLD case 2, episode #1), UA9-I, UA22-N,UA24-I

Cluster 5: UA1-U, UA27-H (PTLD case 3),NKB4-L, UA15-N, NKB6-H, NKB9-H,C3-P, UA30-I, UA12-U, UA20-P, UA21-P(N, control negative; P, control positive; U,undetectable; L, low; I, Intermediate; H,high).

Table 2 shows some order to the clustering ofgenes according to viral load measurements.Cluster 2 as shown is entirely made up of patientswith low or undetectable viral load. The majorcontributors to clusters 1, 4, and 5 are patientswith high or intermediate levels of viral load.

Supervised analysis of multiple group comparison

We performed four comparisons: EBV-seroposi-tive (P) vs. EBV-seronegative controls (N); sus-tained high plus intermediate high EBV load (HI)vs. low plus undetectable loads (LU); sustainedhigh plus intermediate high EBV load (HI) vs.seropositive controls (P) and low plus undetect-able loads (LU) vs. seropositive controls (P).Given an FDR level = 0.1, the comparisons

of HI vs. P and LU vs. P identified 24 significantexpression genes. The other two comparisonsHI vs. LU and P vs. N identified 54 and 37significant expressed genes, respectively. Tables 3and 4 show the genes with the FDR £ 0.1 for thecomparisons HI vs. P and LU vs. P, respectively.

Fig. 1. Results of cluster analysis of 563 gene expression profiling. Expression of host genes by CD19+ B lymphocytes in asample of 33 study patients, including controls (34 specimens). The dendrogram shows clusters of 563 variable genes. Fiveclusters can be identified, numbered from left to right in the figure above. The names of the genes are not visible at the extremeright of the figure because of the large number. Table 2 shows the categories of subjects within each cluster. Upregulated genesare shown in red and derepressed genes are shown in blue.

Table 2. Clustering of significant host genes according to levels of viral loads

Categories of patients Cluster 1 (%) Cluster 2 (%) Cluster 3 (%) Cluster 4 (%) Cluster 5 (%)

Low–undetectable EBV load 14.3 35.7 14.3 14.3 21.4High–intermediate EBV load 23.1 0 15.4 30.8 30.8Healthy controls–EBV-seropositive 0 0 33.3 0 66.7Healthy controls–EBV-seronegative 25 0 25 25 25

Gene expression and EBV lymphoproliferation

993

The effect of comparing HI vs. LU is shown in adendrogram in Fig. 2. We used RT-PCR tovalidate the microarray data. This is illustratedby Fig. 3a and b, for gene GNLY and EIF1AY.The gene GNLY showed down-regulation in thehigh and intermediate EBV load groups com-pared with the seropositive control group.EIF1AY showed up-regulation in the intermedi-ate EBV load group compared with the seropos-itive control group.

Discussion

Gene expression profiling in EBV-related post-infective fatigue syndrome was recently explored(14). Our preliminary study has provided us withthe opportunity to explore the utility of geneexpression profiling using microarray technologyin the evaluation of EBV-related lymphoprolif-eration after organ transplantation among pedi-atric organ transplant recipients. One importantissue was the need to ensure that adequatequantities of high-quality RNA could beextracted from the peripheral blood CD19+ B

cells obtained from children after transplanta-tion. In this regard, the concern is the variablelymphocyte counts that may be present in somepatients. Our results showed that adequatequality and quantity of RNA could be obtainedfrom the study population. However, it shouldbe noted that in some cases two cycles ofamplification were necessary prior to microarrayanalysis. This raises the concern that suchamplification could artificially increase thechances of detecting genes that are not signifi-cantly over-expressed. This has been shown notto be of significant concern as evidenced by thefact that consistent gene expression profiles havebeen obtained with two or three rounds ofamplification (15–17).Our results demonstrate that there was some

degree of alignment between patients with differ-ent viral loads and gene clusters. With respect tothe specific genes that were expressed orrepressed, our study was not powered to drawdefinitive conclusions at this stage. However,the results have provided useful information in

Table 3. Comparison of patients with high and intermediate viral load and EBV-seropositive controls

Adjustedp-values Unique ID Gene Description

0 Hs.529901 X (inactive)-specific transcript5.16E-53 Hs.99120 DEAD (Asp-Glu-Ala-Asp) box polypeptide 3, Y-linked4.97E-50 Hs.282376 Ribosomal protein S4, Y-linked 1 Major histocompatibility complex, class II, DQ beta 1 ///0 Hs.409934 Major histocompatibility complex, class II, DQ beta 10 Hs.387679 Major histocompatibility complex, class II, DQ alpha 10 Hs.529901 X (inactive)-specific transcript0 Hs.529901 X (inactive)-specific transcript0 Hs.529901 X (inactive)-specific transcript5.56E-28 Hs.461178 Eukaryotic translation initiation factor 1A, Y-linked0 Hs.534322 Major histocompatibility complex, class II, DR beta 10 Hs.529901 X (inactive)-specific transcript5.33E-20 Hs.80358 Smcy homolog, Y-linked (mouse)7.38E-20 – Major histocompatibility complex, class II, DQ alpha 1 /// Major histocompatibility complex, class II, DQ alpha 13.63E-16 Hs.461178 Eukaryotic translation initiation factor 1A, Y-linked2.27E-14 Hs.165856 Ubiquitin specific peptidase 9, Y-linked (fat facets-like, Drosophila)1.87E-14 Hs.592254 Chromosome Y open reading frame 15B0 Hs.534322 Major histocompatibility complex, class II, DR beta 1 /// Major histocompatibility complex, class II, DR beta 37.83E-14 Hs.522863 Chromosome Y open reading frame 15A2.62E-13 Hs.592254 Chromosome Y open reading frame 15B5.39E-06 Hs.461178 Eukaryotic translation initiation factor 1A, Y-linked1.21E-05 Hs.522845 Zinc finger protein, Y-linked0.000426 Hs.592254 Chromosome Y open reading frame 15B0.002795 Hs.99120 DEAD (Asp-Glu-Ala-Asp) box polypeptide 3, Y-linked0.008267 Hs.522863 Chromosome Y open reading frame 15A0.012393 Hs.234642 Aquaporin 3 (Gill blood group)0.013051 Hs.325531 Churchill domain containing 10.053759 Hs.165856 Ubiquitin specific peptidase 9, Y-linked (fat facets-like, Drosophila)0.053759 Hs.90708 Granzyme A (granzyme 1, cytotoxic T lymphocyte-associated serine esterase 3) /// granzyme A (granzyme 1, cytotoxic

T lymphocyte-associated serine esterase 3)0.06394 Hs.409934 Major histocompatibility complex, class II, DQ beta 1 /// Major histocompatibility complex, class II, DQ beta 1

/// Similar to HLA class II histocompatibility antigen, DQ (W1.1) beta chain precursor (DQB1*0501) /// similar toHLA class II histocompatibility antigen, DQ (W1.1) beta chain precursor (DQB1*0501)

0.08342 Hs.105806 Granulysin

Allen et al.

994

three areas. First, we have documented thegenes that are expressed in our populationof transplant recipients without evidence ofEBV lymhoproliferation. This provides usefulreference information that facilitates the inter-pretation of results of future studies on lympho-proliferation.Second, our study results also provide data

that could be of value in the interpretation of thegene expression profiles in future studies of post-transplant conditions that are not related toEBV. While it is not our intent to make extensivespeculation about clinical correlates at this stage,the data that we have obtained will be enteredinto our data library to have a repository ofbackground reference data for future studies ondifferent conditions in transplant populations.Third, an important feature of the interaction

between EBV and the immune system is theinduction of a TH2 cytokine/chemokine profilethat is favorable to viral replication (e.g., inhibi-tion of interferon-c production resulting in lessCTL activity). To evade the immune system,EBV alters the balance between TH1 and TH2response profiles (18, 19). These profiles may bealigned with genes that are expressed by EBV-infected B cells (19, 20). This underlying premisehelps to guide our focus of research.

In the above context, our exploratory analysesfor differential expression of genes show genesthat have a plausible biologic role in the patho-genesis of EBV lymphoproliferation and whichare aligned with the cytotoxic lymphocyte activ-ity. Examples of the genes that we identifiedinclude, but are not limited to CCL5 and SLAMfamily member 7. In this regard, EBV infection isable to induce B cells to up-regulate theirconstitutive expression of various chemokines,such as CCL5 (Rantes), which are known toattract TH1 cells and activated cytotoxic T cellsvia the receptor CCR5 (19). The SLAM familyhas a crucial role to play in the control of EBVinfection. This is illustrated by the fact thatX-linked lymphoproliferative disease is an oftenfatal entity that is characterized by an inability tocontrol EBV (21–23). This condition results frommutations in the SH2D1A gene, which encodesthe adaptor protein signaling lymphocytic acti-vation molecule-associated protein. The latter isexpressed in T lymphocytes and NK cells, whileSLAM is expressed by activated lymphocytesincluding EBV-infected B lymphocytes (24).Future studies with sufficient numbers of patientswith EBV lymphoproliferation and PTLD willshed further light on the roles, if any, of these andrelated gene profiles in predicting the likelihood

Table 4. Comparison of patients with low and undetectable viral load and EBV-seropositive controls

Adjustedp-Value Unique ID Gene description

0 Hs.409934 Major histocompatibility complex, class II, DQ beta 1 /// Major histocompatibility complex, class II, DQ beta 10 Hs.555872 Major histocompatibility complex, class II, DQ alpha 10 Hs.555872 Major histocompatibility complex, class II, DQ alpha 12.50E-06 Hs.567449 Immunoglobulin heavy locus /// immunoglobulin heavy constant gamma 1 (G1m marker) /// immunoglobulin heavy

constant gamma 2 (G2m marker) /// immunoglobulin heavy constant gamma 3 (G3m marker) /// immunoglobulinheavy constant mu

1.72E-05 Hs.558339 Immunoglobulin heavy constant alpha 1 /// immunoglobulin heavy constant alpha 2 (A2m marker) /// hypotheticalprotein MGC27165

0.000155 Hs.517265 SLAM family member 70.001184 Hs.325531 Churchill domain containing 10.001093 Hs.398636 Hemoglobin, alpha 2 /// hemoglobin, alpha 20.001184 Hs.409934 Major histocompatibility complex, class II, DQ beta 1 /// Major histocompatibility complex, class II, DQ beta 10.004694 Hs.520421 PERP, TP53 apoptosis effector0.006544 Hs.128686 Nucleobindin 20.011506 Hs.348264 Granzyme H (cathepsin G-like 2, protein h-CCPX) /// granzyme H (cathepsin G-like 2, protein h-CCPX)0.011158 Hs.50749 Tripartite motif-containing 40.025697 Hs.234642 Aquaporin 30.025697 Hs.512656 Phosphoserine phosphatase0.025697 Hs.114198 Mitogen-activated protein kinase kinase 50.039894 Hs.508880 T cell receptor alpha locus /// T cell receptor delta locus0.043639 Hs.105806 Granulysin0.043639 Hs.10306 Natural killer cell group 7 sequence0.062259 Hs.558660 Sterile alpha motif domain containing 30.044708 Hs.514821 Chemokine (C-C motif) ligand 50.065761 Hs.381568 Immunoglobulin J polypeptide, linker protein for immunoglobulin alpha And mu polypeptides0.065761 Hs.2556 Tumor necrosis factor receptor superfamily, member 170.092519 Hs.517265 SLAM family member 7

Gene expression and EBV lymphoproliferation

995

that patientswith elevated viral loadswill progresstoPTLD.Within groupsof patientswithhighviralload, one might be able to identify those are mostlikely to developPTLDbyvirtue of the presence ofspecific gene markers. Ultimately, such markerscould then be identified using technologies thatcan be applied to clinical laboratories.We acknowledge that for many studies using

technology similar to ours, there are somelimitations. An alternative approach would havebeen to focus on CD19+ EBV-infected B cells toreduce the ‘‘background noise’’ created by non-infected B cells. This approach would be techni-cally more cumbersome. However, by comparingthe expression profiles from CD19+ B cells frominfected vs. non-infected individuals, we are able

to delineate the effects attributable to EBVinfection. We also noted that some ‘‘backgroundnoise’’ was created by small numbers of contam-inating T and NK cells.Our study was not meant to be all-encompass-

ing. In future studies, we are addressing a widerrepertoire of cells, including CD4+ and CD8+T cells. In addition, the numbers of PTLDpatients were too few to draw definitive conclu-sions about PTLD. As such, larger numbers ofpatients with PTLD would have enhanced ourstudy; this is currently the focus of subsequentstudies that will be enhanced through multicentercollaboration.In interpreting the results of studies such as

ours, it is important to consider the effects of

Fig. 2. Expression of host genes by CD19+B lymphocytes in a sample of 27 immunosuppressed patients (transplant recipients)with high/intermediate vs. low/undetectable Epstein–Barr Viral loads. The above dendrogram shows differentially expressedprobe sets (genes) selected at a false discovery rate of 0.1. High and intermediate viral load categories were grouped together,while low and undetectable were also grouped. For the most part, the grouping of the subjects is aligned with the way that thedifferentially expressed genes are clustered. Upregulated genes are shown in red and derepressed genes are shown in blue.

Allen et al.

996

known as well as unknown confounders. Ideally,one would like to tease of the effects of severalvariables, including age, cell counts, types oforgan transplanted and the types of specificimmunosuppression. While our exploratory anal-yses did not reveal dominant effects attributableto these variables in our study, based on ourresearch objectives, a more in depth study ofthese is warranted in the future. With respect tothe effects of immunosuppression, futureresearch would include larger numbers ofpatients who have received distinctly differentimmunosuppressive agents.In summary, this is a proof-of-concept study

that will facilitate further studies that examinedifferences in gene expression in PTLD and thespectrum of EBV lymphoproliferation. Giventhat our sample size is small, further validation isrequired to handle the effects of confounders thatare difficult to control in situations of low samplesizes. This notwithstanding, this study providesinformation that supports plans for subsequentstudies to examine the role of gene expressionprofiling in predicting the likelihood that patientsat risk of EBV lymphoproliferation will developPTLD. To this end, in follow-up to this study,

additional studies are in progress including theassembling of a larger prospective cohort ofpatients including those with sustained elevationsof EBV load and PTLD.

Acknowledgements

This study was facilitated in part by a grant from theCanadian Institutes ofHealth Research (grant #MOP62936).

References

1. Henle W, Diehl V, Kohn G, Zur Hausen H, Henle G.

Herpes-type virus and chromosome marker in normal leuko-

cytes after growth with irradiated Burkitts cells. Science 1967:

157: 1064–1065.

2. PopeJH,HorneMK,ScottW.Transformationof foetalhuman

leukocytes in vitro by filtrates of a human leukaemic cell line

containing herpes-like virus. Int J Cancer 1968: 3: 857–866.

3. Petti L, Kieff E. A sixth Epstein–Barr virus nuclear antigen

(EBNA 3B) is expressed in latently infected growth-trans-

forming lymphocytes. J Virol 1988: 62: 2173–2178.

4. Ho M. Risk factors and pathogenesis of posttransplant

lymphoproliferative disorders. Transplant Proc 1995: 27:

38–40.

5. Allen U, Hebert D, Petric M, et al. Utility of semiquanti-

tative polymerase chain reaction for Epstein–Barr virus to

measure virus load in pediatric organ transplant recipients with

and without posttransplant lymphoproliferative disease. Clin

Infect Dis 2001: 33: 145–150.

6. Green M, Cacciarelli TV, Mazariegos GV, et al. Serial

measurement of Epstein–Barr viral load in peripheral blood in

pediatric liver transplant recipients during treatment for

posttransplant lymphoproliferative disease. Transplantation

1998: 66: 1641–1644.

7. Preiksaitis JK, Keay S. Diagnosis and management of post-

transplant lymphoproliferative disorder in solid-organ trans-

plant recipients. CID 2001: 33: S38–S46.

8. QIAGEN. Available at: http://www.qiagen.com/RNA (Acces-

sed September 14, 2008).

9. Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B,

Speed TP. Summaries of Affymetrix GeneChip probe level

data. Nucleic Acids Res 2003: 31: 1–8.

10. R statistical package. http://www.r-project.org/ (Accessed

September 14, 2008).

11. Jain N, Thatte J, Braciale T, Ley K, O�Connell M, Lell

JK. Local pooled error test for identifying differentially

expressed genes with a small number of replicated microarrays.

Bioinformatics 2003: 19: 1945–1951.

12. Benjamini Y, Hochberg Y. Controlling the false discovery

rate: A practical and powerful approach to multiple testing.

Journal of the Royal Statistical Society B 1995: 85: 289–300.

13. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster

analysis and display of genome-wide expression patterns. Proc

Natl Acad Sci USA 1998: 95: 14863–14868.

14. Cameron B, Galbraith S, Zhang Y, et al. Gene expression

correlates of postinfective fatigue syndrome after infectious

mononucleosis. J Infect Dis 2007: 196: 56–66.

15. Luzzi V, Mahadevappa M, Raja R, Warrington JA,

Watson MA. Accurate and reproducible gene expression

profiles from laser capture microdissection, transcript amplifi-

cation, and high density oligonucleotide microarray analysis.

J Mol Diagn 2003: 5: 9–14.

16. Feldman AJ, Costouros NG, Wang E, et al. Advantages of

mRNA amplification for microarray analysis. BioTechniques

2002: 33: 906–914.

(a)

(b)

Fig. 3. Fold changes of the genes GNLY (a) and EIF1AY(b) in expression data and qPCR data.

Gene expression and EBV lymphoproliferation

997

17. Polacek DC, Passerini AG, Shi C, et al. Fidelity and

enhanced sensitivity of different transcription profiles following

linear amplification of nanogram amounts of endothelial

mRNA. Physiol Genomics 2003: 13: 147–156.

18. Steigerwald-Mullen P, Kurilla MG, Braciale TJ. Type 2

cytokines predominate in the human CD4+ T-lymphocyte

response to Epstein–Barr Virus nuclear antigen 1. J Virol 2000:

74: 6748–6759.

19. Nakayama T, Hieshima K, Nagakubo D, et al. Selective

induction of Th2-attracting chemokines CCL17 and CCL22 in

human B cells by latent membrane protein 1 of Epstein–Barr

Virus. J Virol, 2004: 78: 1665–1674.

20. Nakayama T, Fujisawa R, Izawa D, et al. Human B cells

immortalized with Epstein–Barr virus upregulate CCR6 and

CCR10 and downregulate CXCR4 and CXCR5. J Virol 2002:

76: 3072–3077.

21. Engel P, Eck MJ, Terhorst C. The SAP and SLAM families

in immune responses and X-linked lymphoproliferative disease.

Nat Rev Immunol 2003: 3: 813–821.

22. Latour S, Veillette A. Molecular and immunological basis of

X-linked lymphoproliferative disease. Immunol Rev 2003: 192:

212–224.

23. Nagy N, Cerboni C, Mattsson K, et al. SH2D1A and SLAM

protein expression in human lymphocytes and derived cell lines.

Int J Cancer 2000: 88: 439–447.

24. Henning G, Kraft MS, Derfuss T, et al. Signaling lympho-

cytic activation molecule (SLAM) regulates T cellular

cytotoxicity. Eur J Immunol 2001: 31: 2741–2750.

Allen et al.

998