FEATURED NEW INVESTIGATORAssessment of the translational value of mouse lupusmodels using clinically relevant biomarkers

ANDREW T. BENDER, YIN WU, QIONGFANG CAO, YUEYUN DING, JUDITH OESTREICHER,MELINDA GENEST, SANDEEP AKARE, SALLY T. ISHIZAKA, and MATTHEW F. MACKEY

ANDOVER, MASS

Andrew T. Bender, PhD, is a SePharmacology division at EDevelopment Institute.From Eisai Inc., Andover, MA.

Submitted for publication July 16, 2

17, 2013; accepted for publication

Lupus is an autoimmune disease with a poorly understood etiology that manifestswith a diverse pathology. This heterogeneity has been a challenge to clinical drugdevelopment efforts. A relateddifficulty is the uncertain translational power of animalmodels used for evaluating potential drug targets and candidate therapeutics,because it is unlikely that any 1 preclinical model will recapitulate the spectrum ofhuman disease. Therefore, multiple models, along with an understanding of the im-mune mechanisms that drive them, are necessary if we are to use them to identifyvalid drug targets and evaluate candidate therapies successfully. To this end, wehave characterized several different mouse lupus models and report their differ-ences with respect to biomarkers and symptoms that are representative of the hu-man disease. We compared the pristane-induced mouse lupus disease modelusing 3 different strains (DBA/1, SJL, BALB/c), and the spontaneous NZB x NZWF1(NZB/W) mousemodel. We show that themodels differ significantly in their autoan-tibody profiles, disease manifestations such as nephritis and arthritis, and expressionof type I interferon-regulated genes. Similar to the NZB/W model, pristane-induceddisease in SJL mice manifests with nephritis and proteinuria, whereas the pristane-treated DBA/1 mice develop arthritis and an interferon-driven gene signature thatclosely resembles that in human patients. The elucidation of each model’s strengthsand the identification of translatable biomarkers yields insight for basic lupusresearch and drug development, and should assist in the proper selection of modelsfor evaluating candidate targets and therapeutic strategies. (Translational Research2014;163:515–532)

Abbreviations: BUN ¼ blood urea nitrogen; cDNA ¼ complementary DNA; ELISA ¼ enzyme-linked immunosorbent assay; FC ¼ fold-change; IFN ¼ interferon; Ig ¼ immunoglobulin;MPO ¼ myeloperoxidase; NZB/W ¼ NZB x NZW F1; PBMC ¼ peripheral blood mononuclearcell; PBS ¼ phosphate-buffered saline; RiboP ¼ ribosomal phosphoprotein P; qPCR ¼ quanti-tative polymerase chain reaction; SLE ¼ systemic lupus erythematosus; TLDA ¼ Taqman low-density array; TMPD (pristane) ¼ 2,6,10,14-tetramethylpentadecane; UACR ¼ urinary albumin-to-creatinine ratio

nior Scientist in the PreclinicalMD Serono Research and

013; revision submitted December

January 3, 2014.

Reprint requests: Sally T. Ishizaka, PhD, Eisai, Inc., 4 Corporate

Drive, Andover, MA 01810; e-mail: sally_ishizaka@eisai.com.

1931-5244/$ - see front matter

� 2014 Mosby, Inc. All rights reserved.

http://dx.doi.org/10.1016/j.trsl.2014.01.003

515

AT A GLANCE COMMENTARY

Bender AT, et al.

Background

Systemic lupus erythematosus remains an incom-

pletely addressed medical need of very heteroge-

neous presentation. As the genetics and

pathology of the disease are elucidated, appro-

priate methods to translate from murine preclinical

models to testing in appropriate patient subpopula-

tions are needed.

Translational Significance

Using 4 different mouse lupus models, we exam-

ined clinically relevant symptoms and markers of

disease, including proteinuria, arthritis, autoanti-

bodies, and interferon gene signature, and found

they differed significantly among models. The re-

sults should assist in choosing the correct model

to validate targets, establish preclinical proof of

concept for different disease populations, and to

identify biomarkers for clinical use.

Translational Research516 Bender et al June 2014

Lupus is an autoimmune disease that can affectnumerous organ systems with varying severity,including the skin, musculoskeletal system, kidneys,and central nervous system. The disease can reducequality of life significantly and is fatal in some cases.Lupus is not only variable in symptomatology but alsoin its disease pathogenesis and etiology.1 The etiologyof the disease is poorly understood, but is likely a resultof a complex interplay between environmental and ge-netic factors.2 Disease pathogenesis involves the gener-ation of autoantibodies and their subsequent depositionin end organs. Although a wide variety of different auto-antibody reactivities has been found in patients withlupus, and they have served as biomarkers, correlatingthem with disease subsets or mechanisms has beencomplicated. Exploring the causal link of autoantibodyspecificity and disease severity and outcome remainsan active area of research.3,4 Given the significantheterogeneity in human lupus etiology, pathology, andsymptomatology, it is unlikely that a single animalmodel will recapitulate all the different patientsubsets. Therefore, a variety of animal models may beneeded to represent more fully the lupus patientpopulation to enable accurate evaluation of candidatelupus drugs. Understanding the diverse models alsofacilitates preclinical analysis of potential biomarkerand pharmacodynamic readouts in preparation for theclinic.

A variety of mouse lupus models have beendescribed,5 including those with a genetic susceptibilityto spontaneous disease development (ie, NZB/W,BXSB-Yaa, and MRL/lpr) and those in which diseaseis chemically induced (ie, pristane). In the NZB/Wmodel, lupus-like disease develops spontaneously as aresult of a combination of at least 3 genetic lociaffecting immune activation, apoptosis, and end organsusceptibility to damage.6 A newer, less commonlyused model is the pristane-induced model, which in-volves intraperitoneal injection of the hydrocarbon2,6,10,14-tetramethylpentadecane (TMPD, alsocommonly known as pristane) to trigger lupus-like dis-ease development.7 The pristane model can be per-formed in a variety of mouse strains, and interestingvariations in the lupus disease characteristics andseverity have been documented across different strains.8

There are few available efficacious lupus treatments,and unmet need remains high. A critical step in thedevelopment of new treatments is the evaluation ofcandidate drugs and targets in animal lupus models.The NZB/W and MRL/lpr models have served asthe primary preclinical drug evaluation models,5,9

whereas the pristane model has not been usedroutinely for this purpose. The selection of anappropriate mouse model for evaluating a candidatelupus drug is crucial given the time-intensive nature ofthese lengthy studies and the resource investment asso-ciated with running lengthy lupus clinical trials. Havingmouse lupus models that are characterized with regardto etiology, pathogenesis, and translatable readouts fordisease symptoms and drug impact would helpimmensely in determining how these models relate tothe human lupus condition and would allow selectionof appropriate models for target evaluation, drug evalu-ation, and biomarker discovery studies.To address this issue, we characterized the pathogen-

esis and disease manifestations of several differentmouse lupus models and explored clinically relevantbiomarkers that increase their translational value. Spe-cifically, we compared the NZB/W lupus model andthe pristane model performed in the DBA/1, BALB/c,and SJL mouse strains. Our findings indicate that themodels differ in the timing for disease development,pathogenesis of the disease, and the end organ manifes-tations. Furthermore, we have used advanced biomarkermethods such as autoantibody profiling, bioluminescentimaging, and quantitative polymerase chain reaction(qPCR) analysis of interferon (IFN)-driven gene expres-sion that may elucidate further the translatability of themodels. This report increases the understanding andtranslational value of mouse lupus models describedin this study and should aid in model selection for the

Translational ResearchVolume 163, Number 6 Bender et al 517

purposes of lupus basic research and development ofnew therapies.

MATERIALS AND METHODS

Mouse lupus models. All animal studies were per-formed with the approval of the Eisai InstitutionalAnimal Care and Use Committee in an AAALAC-accredited animal facility. Mice were group-housedunder controlled conditions with a constanttemperature (21�C–23�C) and humidity (30%–50%), a10-hour/14-hour light/dark cycle, and ad libitumaccess to water and standard pelleted food. FemaleNZB/W F1/J mice were purchased from Jackson Labs(Bar Harbor, Maine). For pristane model studies,female DBA/1 mice were purchased from HarlanLaboratories (Indianapolis, Ind) or Jackson Labs (BarHarbor, Maine), BALB/c females were from CharlesRiver (Wilmington, Mass), and female SJL mice werefrom Jackson Labs. Lupuslike disease developedspontaneously in the NZB/W mice but was triggeredin the other strains by an intraperitoneal injection of0.5 mL pristane (Sigma, St. Louis, Mo) at 11–12 weeks of age. Intraperitoneal injection of 0.5 mLphosphate-buffered saline (PBS) was performed togenerate age- and sex-matched control mice withineach study. At the conclusion of studies, mice wereeuthanized by CO2 asphyxiation, and blood wascollected for plasma isolation.

Autoantibody measurements. Enzyme-linked immuno-sorbent assay. Anti-dsDNA, ribosomal phosphoproteinP (RiboP), and histone titers were evaluated by a stan-dard enzyme-linked immunosorbent assay (ELISA)approach. Briefly, 96-well EIA/RIA ELISA plates(Corning) were coated with 100 mL diluted antigen inPBS for 90 minutes at room temperature with (final con-centrations): 10 mg/mL calf thymus dsDNA (Sigma), 5U/mL RiboP (Immunovision), and 5 mg/mL histones(Immunovision). Plates were washed with PBS/0.05%Tween 20 (washing buffer) and blocked overnightwith PBS/1% bovine serum albumin (blocking buffer)at 4�C. Plates were washed, and mouse plasma sampleswere diluted in blocking buffer (ranging from 1:25 to1:10,000, depending on the model and the antigen),added to wells in 100-mL volumes per well, and incu-bated for 90 minutes at room temperature. Plates werethen washed, 100 mL antimouse-immunoglobulin (Ig)G-HRP (Southern Biotech) diluted in PBS/1% bovineserum albumin/0.05% Tween at 1:50,000 was addedto each well, and plates were incubated for 90 minutesat room temperature. Next, plates were washed and100 mL of a 1:1 mix of substrate components from theOptEIATMB substrate kit (BD Biosciences) was addedto the wells. Plates were incubated at room temperatureand, after sufficient color development, the reaction was

stopped by adding 100 mL 0.18M sulfuric acid solution.Plates were read by spectrophotometry at 450 nm. Astandard curve was generated for each autoantibody us-ing a pooled plasma stock isolated from diseased MRL-lpr/lpr mice. Dilutions of the plasma were made toconstruct the curve, and arbitrary units were assignedto allow for quantitation of autoantibody titers.

Autoantigen microarray profiling. Autoantibodyprofiling was performed by the University of TexasSouthwestern Microarray core facility as described pre-viously.10 Briefly, diluted mouse plasma samples wereincubated on slides spotted with 70 different antigens.Autoantibodies bound to the antigens were detected us-ing a Cy3-tagged secondary anti-mouse IgG. Normali-zation was performed according to the total IgG levelsfor each plasma sample. A fold change in autoantibodylevels was calculated for the pristanemice by comparingage- and sex-matched PBS-treated nondiseased micewith pristane-diseased mice. For the NZB/W studies, afold change was calculated by comparing 2-month-oldpre-disease mice with mice with late-stage disease (9–12 months of age). The reactivities were then ranked ac-cording to fold change within each study. A heat mapshowing the rankings of the autoantibodies withineach study was then constructed, with the autoanti-bodies having the highest expression assigned the high-est number and the autoantibodies with least expressionassigned the lowest number.

Nephritis measurements. The urinary albumin-to-creatinine ratio. Urine was collected from individualSJL mice in metabolic cages for 18 hours. From allother mice, urine was collected manually at singletime points in the morning. The urinary albumin-to-creatinine ratio (UACR) was determined for eachanimal as a measure of proteinuria and kidneyfunction. UACR was calculated as the ratio ofmilligrams of albumin per gram of creatinine perdeciliter of urine. Albumin levels in the urine sampleswere determined using a custom sandwich ELISAprotocol developed in-house, using an antimousealbumin antibody set (Bethyl Labs, Montgomery, Tex)that included a coating antibody and an HRP-taggedsecondary for detection. Creatinine levels weredetermined using a commercial creatinine assay kit(Cayman Chemical, Ann Arbor, Mich).

Histology. Mice were euthanized and the kidneyswere fixed in 10% formalin. The formalin-fixed tissuewas paraffin embedded, sectioned, and stained withhematoxylin-eosin stain at HistoTox Labs (Boulder,Colo). Kidney slides were evaluated in a blinded fashionby a trained veterinary pathologist, and glomerulone-phritis and overall kidney pathology were characterized.Key points of the grading system were as follows: grade0/1, normal to below threshold change; grade 2,

Translational Research518 Bender et al June 2014

minimal change characterized by glomerular increasedeosinophilic matrix, thickening of capillary basementmembrane or luminal capillary thrombus-like material,often affecting only a few glomeruli; grade 3, moderateto marked change of much greater severity than grade 2,often affecting the majority of glomeruli but lackingglomerulosclerosis; and grade 4, severe changeaffecting the majority of glomeruli and often accompa-nied by glomerulosclerosis, marked tubular dilatation,and protein casts.

Blood urea nitrogen. Blood urea nitrogen (BUN) wasmeasured in plasma samples using the Heska Dri-Chem 7000 Veterinary Blood Chemistry Analyzer(Loveland, Colo). The analyzer was used in ManualSpotting mode and 10 mL plasma was spotted ontoeach slide.

Arthritis measurements. Clinical scoring. Mice werescored for clinical arthritis by 2 observers who wereblinded to the group assignments. Arthritis was notedas an inflammation and swelling of the paws, and dis-ease scores were assigned in a blinded analysis to quan-titate the arthritis similar to a method that has beendescribed previously for other mouse arthritis models.11

Each paw was scored on a scale of 0–4, yielding amaximum score of 16 per mouse.

Luminol-based bioluminescence imaging. Mice wereimaged for myeloperoxidase (MPO) activity in paws us-ing a technique described previously.12 Briefly, micewere injected with 200 mg/kg luminol sodium salt solu-tion (Sigma) intraperitoneally and then anesthetized us-ing 1.5% isoflurane. Twelve minutes after the luminolinjection, mice were imaged for bioluminescence usingthe IVIS Spectrum imaging system (Caliper Lifescien-ces, Hopkington, Mass). Bioluminescence in the hindpaws was quantitated using the Living Image softwareprogram.

IFN gene signature analysis. The expression of type IIFN-regulated genes was measured in mouse orhuman whole blood or in isolated human peripheralblood mononuclear cells (PBMCs) stimulated inculture.

Mouse TaqMan low-density array. The IFN gene signa-ture for whole mouse blood was measured by qPCR at5 months after pristane injection. Briefly, mice wereeuthanized, blood was collected via the vena cava, and100mLof bloodwas preserved in tubes containingRNA-later (Ambion, Austin Tex). Total RNAwas isolated us-ing the Mouse RiboPure Blood RNA Isolation Kit(Ambion). RNA concentrations were determined usinga NanoDrop ND-1000 spectrophotometer (Thermo Sci-entific, Waltham, Mass). First-strand cDNAwas synthe-sized from 100 ng total RNA using the High CapacityRNA-to-cDNA Master Mix (Applied Biosystems, Fos-

ter City, Calif). After reverse transcription, cDNA wasdiluted with nuclease-free water and mixed with GeneExpression Master Mix (Applied Biosystems). Themixture was then applied to a custom TaqMan low-density array (TLDA) manufactured by Applied Bio-systems, and qPCR was performed on the ABI7900HT Fast Real-Time PCR System (Applied Bio-systems). Raw data were collected using RQ Manager1.2.1 (Applied Biosystems) and were analyzed usingGeneData Analyst 2.2 software (GeneData). The mouseTLDA panel, described in Table I, contained 36 targetgenes and 3 housekeeping genes for normalization.The housekeeping gene Hprt1 was chosen for normali-zation based on a low coefficient of variation. Relativequantification of the target genes was performed tocalculate a fold change for each diseased mouse relativeto the mean of the nondiseased mouse group receivingintraperitoneal PBS injection only. For mouse TLDAdata, the cutoff used for significance was a P value ofless than 0.05 by Student’s t test, with an fold-change(FC) of more than 1.5. Genes with very low expressionwere eliminated and are shaded in gray in Table I. An‘‘IFN score’’ was calculated subsequently for eachmouse and is the median fold change of all detectablegenes in the pristane-injected mice. When fold changesare listed in tabular form, the FC column shows themeanfold change for the pristane-treated individual animals.

Human TLDA. For validation of the IFN-regulatedgenes, human blood was collected from consentinghealthy female donors into heparin anticoagulant-treated tubes. PBMCs were isolated using Histopaque1077 (Sigma) and cultured in RPMI 1640 1 10% fetalbovine serum. Cells were treated for 20 hours with amixture of IFN subtypes (1000 U/mL of each subtype):IFN-alpha subtypes A, B2, C, D, K, F, G, H2, I, J1, 4b,WA, and IFN-b (all obtained from PBL InterferonSource, Piscataway NJ). After culture, total RNA wasisolated from the cells and processed as described earlierfor analysis on a TLDA array, with human versions of asubset of genes assayed in the mouse TLDA array. Forthe human TLDA array analysis, genes from the mousearray that had very low expression were eliminated, aswere any genes that did not have exactlymatching ortho-logs (IFI204). Gene expression levels were determinedrelative to untreated PBMCs collected after 3 hours ofculture. The cutoff for significance was a P value lessthan 0.05 by t test, with a FC of more than 2. Themean of four housekeeping genes (GAPDH, GUSB,ZNF592, and PGK1) was used for normalization.

Systemic lupus erythematosus patient sampleanalysis. ELISAs. Blood samples were collected from20 consenting female patients with systemic lupus ery-thematosus (SLE) and 20 female healthy donors by Bio-reclamation LLC (Westbury, NY) by trained personnel

Table I. Interferon gene signature TLDA

Gene symbol Taqman ID, mouse Taqman ID, human Gene name

18S Hs99999901_s1 Hs99999901_s1 Eukaryotic 18S rRNABst2 Mm01609165_g1 Hs00171632_m1 Bone marrow stromal cell antigen 2Ccl2 Mm00441243_g1 NA Chemokine (C-C motif) ligand 2Ccr2 Mm00438270_m1 Hs00704702_s1 Chemokine (C-C motif) receptor 2Cd300e Mm00468131_m1 NA CD300e antigenCxcl10 Mm00445235_m1 NA Chemokine (C-X-C motif) ligand 10Cxcr5 Mm00432086_m1 NA Chemokine (C-X-C motif) receptor 5Elane Mm00469310_m1 NA Elastase, neutrophil expressedEpsti1 Mm00712734_m1 Hs01566789_m1 Epithelial stromal interaction 1 (breast)Fcgr1 Mm00438874_m1 NA Fc receptor, IgG, high affinity IFpr1 Mm00442803_s1 NA Formyl peptide receptor 1Gapdh Mm99999915_g1 Hs02758991_g1 Glyceraldehyde-3-phosphate dehydrogenaseHprt Mm00446968_m1 NA Hypoxanthine guanine phosphoribosyltransferaseIfi202 b Mm00839397_m1 NA Interferon-activated gene 202BIfi204 Mm00492602_m1 NA Interferon-activated gene 204Ifi35 Mm00510329_m1 Hs00413458_m1 Interferon-induced protein 35Ifi44 Mm00505670_m1 NA interferon-induced protein 44Ifit1 Mm00515153_m1 Hs01911452_s1 Interferon-induced protein with tetratricopeptide repeats 1Irf7 Mm00516788_m1 Hs01014809_g1 Interferon regulatory factor 7Isg15 Mm01705338_s1 Hs01921425_s1 ISG15 ubiquitinlike modifierIsg20 Mm00469585_m1 Hs00158122_m1 Interferon-stimulated proteinLy6e Mm01200460_g1 Hs00158942_m1 Lymphocyte antigen 6 complex, locus EMmp8 Mm00439509_m1 NA Matrix metallopeptidase 8Mmp9 Mm00442991_m1 Hs00234579_m1 Matrix metallopeptidase 9Mpo Mm00447886_m1 NA MyeloperoxidaseMs4a6c Mm00459296_m1 NA Membrane-spanning 4-domain, subfamily A, member 6CMx1 Mm00487796_m1 NA Myxovirus (influenza virus) resistance 1Oas3 Mm00460944_m1 Hs00196324_m1 2-5 Oligoadenylate synthetase 3Oasl2 Mm00496187_m1 NA 2-5 Oligoadenylate synthetaselike 2Ppia Mm02342430_g1 NA Peptidylprolyl isomerase A (cyclophilin A)Rsad2 Mm00491265_m1 Hs00369813_m1 Radical S-adenosyl methionine domain containing 2Stat1 Mm00439531_m1 Hs01013996_m1 Signal transducer and activator of transcription 1Tnfsf13 b Mm00446347_m1 NA Tumor necrosis factor (ligand) superfamily, member 13bTreml4 Mm00553947_m1 NA Triggering receptor expressed on myeloid cellslike 4Usp18 Mm00449455_m1 Hs00276441_m1 Ubiquitin-specific peptidase 18Xaf1 Mm01248390_m1 Hs00213882_m1 XIAP associated factor 1

Abbreviations: ID, identification; Ig, immunoglobulin; NA, not applicable; rRNA, ribosomal RNA; TLDA, Taqman low-density array.

Translational ResearchVolume 163, Number 6 Bender et al 519

at U.S. Food and Drug Administration-licensed and -in-spected collection centers according to institutional re-view board approved procedures. Blood samples weredrawn into heparinized BDVacutainers (Becton Dickin-son, Franklin Lakes, NJ) and also into PAXgene BloodRNA tubes (Beckton Dickinson). The samples wereshipped overnight before processing and analysis. Theanticoagulated whole blood was subjected to centrifu-gation for isolation of plasma. Plasma samples wereanalyzed by ELISA for titers of autoantibodies asdescribed earlier, with the exception that an antihumanIgG (Bethyl Labs) was used for autoantibody detection,and raw OD values are reported because no standardcurve was generated in the human assays.

Human microarray analysis. DNA microarray geneexpression analysis was performed on the SLE patientand healthy donor blood samples collected into the

PAXgene Blood RNA tubes. RNA was isolated usingthe PAXgene miRNA Isolation Kit (Qiagen, German-town, Md), and globin reduction was performed usingthe GLOBINclear Kit (Life Technologies). The sense-strand cDNAwas generated from total RNA using Am-bion WT Expression kit (Life Technologies), andmicroarray analysiswas carried out using theAffymetrixhuman exon array (HuEx-1_0-st-v2). All chips passedquality control metrics, and data were analyzed usingthe GeneData Expressionist software program. A foldchange was calculated comparing the gene expressionof each patient with SLE with the average of the healthydonor expression level, and a heat map was constructedof selected IFN-regulated genes for individual patientswith SLE. An IFN score was calculated for each patientin the same way as for the mice, and is the median foldchange of the all the genes shown in the heat map.

A

B

Fig 1. (A, B) Time course for development of anti-dsDNA autoanti-

bodies. Serial blood samples were collected over time from mice via

the tail vein for New Zealand black/white (NZB/W) mice (A) and

via the submandibular route for pristane studies (B). The levels of

anti-dsDNA autoantibodies were measured by enzyme-linked immu-

nosorbent (ELISA) assay in the plasma samples. View A is a compos-

ite average of 4 studies and view B is representative of 4 studies. The

mean and standard error of the mean are plotted.

Translational Research520 Bender et al June 2014

Statistics. Statistical analysis was donewith GraphPadPrism. Data are presented as medians with interquartilerange, or as mean6 standard error of the mean, as statedin figure captions. Statistical comparisons on scatter-plots were run using the Mann-Whitney test (whencomparing 2 groups) or Kruskal-Wallis with post test,as described in the figure captions. The analysis ofgene expression data is described in the correspondingMethods section. Correlation of antibody titers withIFN signature was done by Spearman’s rank correlation.

RESULTS

Autoantibody development. Autoantibodies are ahallmark of lupus and are not only diagnostic criteriabut also disease mediators, because they have been re-ported to drive lupus pathogenesis.13 The most widelyclinically monitored autoantibody is anti-dsDNA, andwe measured this reactivity over time to track diseasedevelopment in mouse lupus models. Fig 1 shows thetime course for development of anti-dsDNA titers inNZB/W mice (Fig 1, A) and mice of different strainsinjected with pristane (Fig 1, B). In the NZB/Wmodel, the presence of anti-dsDNA is increasednoticeably at 2–5 months of age and begins to plateauaround 7 months of age. In contrast, anti-dsDNAelevation in the pristane model is noted somewhatearlier, at 3 months after pristane injection, and peaksat 5 months (Fig 1, B) at about a 10-fold lower levelthan observed in NZB/W mice. Among the pristane-induced strains, SJL showed a consistently higher titerof anti-dsDNA than the DBA/1 and BALB/c. The timecourse for anti-dsDNA production is similar for all 3strains after pristane injection.A wide variety of autoantibody reactivities has been

detected in patients with lupus, and there is significantheterogeneity in the reactivity patterns found indifferent individuals.10 Thus, we characterized mouselupus models for their autoantibody profiles to discoverdifferences and correlations to human patients withlupus. Autoantibody profiling was performed using anantigen-coated microarray by the UT Southwestern Mi-croarray Core to determine which reactivities wereelevated with disease in each model (Fig 2, A). As hasbeen reported previously,14 the pristane-diseased miceshowed elevations in autoantibodies against manyRNA-associated antigens, such as U1-snRNP com-plexes and RiboP. These reactivities were elevatedgreatly in both BALB/c and DBA/1 pristane-treatedmice, but were not increased in NZB/W mice, consis-tent with previous findings.15 In contrast, but consistentwith the strong dsDNA reactivity seen in Fig 1, theNZB/W model was characterized by an elevation in au-toantibodies against nuclear antigens such as chromatinand histones. These model-defining clusters are high-

lighted in Fig 2, B, which shows the autoantibodiesmost highly elevated in diseased mice compared withnondiseased mice for each model.To verify the microarray results, the levels of anti-

RiboP autoantibodies in the DBA/1 pristane modeland antihistone autoantibodies in the NZB/W modelwere analyzed by ELISA (Fig 2, C andD, respectively).Both reactivities were elevated significantly in diseasedmice compared with healthy mice, with a high pene-trance. We also found induction of Sm/RNP in the pris-tane mice, with the highest levels in the DBA/1 strain(Supplementary Fig 1). These ELISA results confirmthe microarray data and suggest that measuring specificautoantibody titers by ELISA is a robust method fortracking disease in these models.

Nephritis. One of the early indicators of nephritis andkidney damage in patients with lupus and mouse lupusdisease models is proteinuria, reflecting glomerulardamage. Mechanistically, glomerular permeability isaltered as a result of changes in the porosity and chargeof the filtration membrane, leading to marked leakage ofprotein (especially albumin) into the urine.16 Wetracked proteinuria (albuminuria) in the mouse modelsin longitudinally collected urine samples, and a UACRwas determined. The UACR values for NZB/W miceover time are shown in Fig 3, A. There was

Fig 2. Microarray autoantibody profiling. Blood samples were collected from New Zealand black/white (NZB/

W) mice or pristane model mice and the plasma was assayed for immunoglobulin G (IgG) autoantibody levels

using a microarray containing 70 different antigens. A fold change in each autoantibody was determined for

NZB/W model studies by comparing young healthy mice with older, diseased mice. For the pristane model, a

fold change between healthy phosphate-buffered saline (PBS)-injected mice and pristane-injected diseased

mice was determined. The median fold change for each autoantibody was then determined and the autoantibodies

Translational ResearchVolume 163, Number 6 Bender et al 521

=

Translational Research522 Bender et al June 2014

heterogeneity in the timing of nephritis development,and even at 9 months of age, many animals did notprogress beyond the 2-month baseline. In the pristanemodel, UACR values are presented for 4 months (Fig3, D) after pristane injection. DBA/1 mice showed theleast proteinuria development at this time point,whereas in the BALB/c pristane-treated mice, therewas sporadic proteinuria with few UACR values muchabove the control mice. In the SJL mice, there was ahigher background UACR in uninduced mice, but thisis also the only strain for which a statisticallysignificant proteinuria level was observed at this time.

Kidney histopathology. Histologic analysis of kidneyscan be used to confirm and characterize nephritisseverity.16 In preliminary analysis we examinedhematoxylin-eosin-stained kidney sections frompristane-induced mice of all 3 strains. We found thatpristane-treated SJL mice advance more rapidly thanBALB/c or DBA/1 mice to the higher grades ofhistologic nephritis (Supplementary Fig 2), with allSJL animals reaching grade 2 or higher by around20 weeks after induction.We next performed in-depth histologic assessment of

SJL pristane and NZB/W animals, given that these 2strains in particular demonstrate significant mortality(see Fig 3). In both models, glomerular change associ-ated with kidney damage of mild severity often affectedonly a few glomeruli and had no accompanying tubularchanges, whereas in the kidneys with moderate andmarked severity change, the glomerular involvementwas widespread and was accompanied by tubular baso-philia and/or protein casts in the tubular lumen (seeasterisk in Fig 4). The histology of glomerular changesin NZB/W and SJL pristane nephritis models was com-parable in most aspects. The tubular basophilia accom-panying the moderate and marked glomerular changesin the SJL pristane model was especially widespread.In both models, increased cellularity was noted, aswas thickening of the basement membrane (see Fig 4,black and white arrows, respectively). We also notedthat the SJL strain has a background incidence of focaltubular basophilia. In the SJL pristane model, a few an-imals with marked glomerular changes had severe tubu-lointerstitial changes that included severe tubular

were ranked based on the magnitude of the fold change to gi

bodies were induced most highly. (A) The heat map shows t

cating the autoantibodies with the largest fold change for eac

models as described in the text are expanded to show greater

pristane BALB/c, and 1 pristane SJL study are shown. (C,

(RiboP) in the DBA/1 pristane model at 4–5 months after

the NZB/Wmodel at 9 months of age (D) were verified by en

results are composites of 3 separate studies for both autoantib

Pris, pristane. ***Significant difference with P , 0.001 by

dilatation, tubular protein casts, tubular degeneration/regeneration, and tubulointerstitial inflammation. Incomparison, severity of tubulointerstitial changes,including tubular dilatation and protein casts, was min-imal to moderate in NZB/W model mice, even in ani-mals with marked glomerular changes. In the NZB/Wmodel, in some animals deposition of homogenouseosinophilic material was a dominant feature of glomer-ular change, and was accompanied by occlusion of cap-illaries but with little cellular component. Human lupusnephritis histopathology is very heterogeneous andchanges with time and disease progression. Thus, ef-forts have been made to classify the different stages,and one such classification scheme was published in2003 by the International Society of Nephrology/RenalPathology Society.17 According to this scheme, the mildto moderate lupus nephritis seen in these 2 mousemodels resembles the class III human nephritis criteria,because it is a proliferative glomerulonephritis with lessthan 50% total glomeruli affected. Severe mousenephritis in these studies resembles the class IV humannephritis criteria, with more than 50% total glomeruliaffected. More detailed comparison of the human andmouse nephritis characteristics by immunofluorescenceand electron microscopy would be of interest.

Functional effects of nephritis. We expected kidneyfunction to have a strong impact on mortality of theseanimals, as it does in human patients with lupus. Sur-vival was tracked in several studies, and Kaplan-Meiersurvival curves for NZB/W mice (Fig 3, C) andpristane-treated mice (Fig 3, E) are presented. In thepristane model, strain-dependent differences in thecourse of mortality were observed. BALB/c and DBA/1 mice all survived past 5 months after pristaneinjection, but in most experiments SJL mice showedmortality much earlier, with survival falling at4 months after pristane injection.Mortality was observed in the NZB/W mice begin-

ning after 8 months of age (Fig 3, C), as UACR valuesrose. Survival decreased rapidly and, at 11 months ofage, fell below 50%.Wemeasured BUN, which is an in-dicator of severe kidney damage resulting from signifi-cant loss of kidney function,18 in plasma samples fromthe NZB/W model (Fig 3, B). We found an elevation of

ve a relative comparison to indicate which autoanti-

he ranking for each study, with the color scale indi-

h study in red. (B) The clusters that define different

detail. Results from 3 NZB/W, 3 pristane DBA/1, 2

D) The levels of anti-Ribosomal phosphoprotein P

pristane injection (C) and antihistone antibodies in

zyme-linked immunosorbent assay (ELISA). ELISA

odies. The median and interquartile range are shown.

the Mann-Whitney U test.

Fig 3. (A–E) Proteinuria development and mortality comparison of lupus mouse models. Urine samples were

collected from mice longitudinally. The levels of albumin in the urine were measured by enzyme-linked immu-

nosorbent assay (ELISA), and creatinine concentrations were measured with a kit, using a chemical reaction

for detection. The urinary albumin-to-creatinine ratios (UACRs) were then calculated and plotted as an indication

of proteinuria. UACRs for New Zealand black/white (NZB/W) mice at different ages (A). Pristane model mice at

4 months after pristane injection (D). Phosphate-buffered saline (PBS) mice were age- and sex-matched nondi-

seased mice injected with PBS instead of pristane. ViewA is a composite of 4 experiments and viewD is a compi-

lation of 3 experiments. Blood urea nitrogen (BUN) levels in NZB/W plasma samples at 9–10 months of age were

measured by blood chemistry (B). Results shown are a compilation of 2 experiments. Mice in the different models

were tracked over time for mortality, and the results are plotted as Kaplan-Meier survival curves by age for the

NZB/W model (C) and by time after pristane injection for the pristane model (E). Views C and E are both com-

posites of 3 experiments. *Significantly different from 2-month UACR by Kruskal-Wallis with Dunn’s pairwise

comparison (P , 0.0001). †Significantly different from SJL PBS (P , 0.01). Pris, pristane.

Translational ResearchVolume 163, Number 6 Bender et al 523

BUN in a number of the NZB/W mice, indicating sub-stantial kidney damage in those animals. In addition,in several instances mice were discovered to be mori-bund, and blood samples were collected immediatelybefore euthanasia. Plasma samples from thesemoribundmice all showed very high BUN levels (data not shown),suggesting that kidney failure was likely the cause ofmortality. Thus, BUN, which is also measured in clin-ical practice, can serve as a biomarker for nephritis inthe NZB/W model; and in both the NZB/W and SJLpristane models, early morbidity is a clear, practical

endpoint that could reflect on the ability of treatmentsto block nephritis.

Arthritis. Lupus affects frequently and often severelythe musculoskeletal system and can manifest asarthritis. Thus, a preclinical mouse model that developsarthritis as an aspect of the lupus disease phenotypecould be useful for studying human lupus rheumato-logic complications. To this end, we evaluated arthritisdevelopment in the NZB/W and pristane models, withresults shown in Fig 5. Previous literature reports ofthe pristane model have noted arthritis development

Fig 4. (A–H) Characterization of nephritis. Kidneys from New Zealand black/white (NZB/W) mice and SJL

pristane-diseased mice were evaluated histologically to characterize the nature of the nephritis. Normal histology

of renal glomerulus and cortex for NZB/W mice at 2 months of age (A, B) and histologic changes in glomerulus

and cortex (C, D), respectively, in older, diseased NZB/W nephritic mice. Normal histology of renal glomerulus

and cortex for SJLmice treated with phosphate-buffered saline (E, F) and histologic changes in the glomerulus and

cortex (G, H), respectively, in SJL pristane-diseased nephritic mice. Black arrows indicate regions of increased

cellularity, white arrows point to thickening of the basement membrane, and asterisks highlight examples of

tubular hyaline (protein) casts.

Translational Research524 Bender et al June 2014

after pristane injection in some strains, particularly inDBA/1 mice.14,19 In our studies, arthritis developmentwas not observed in NZB/W mice over their life span.Likewise, arthritis was not detected in SJL mice, but

was seen in both BALB/c and DBA/1 mice onpristane injection (Fig 5, A). Arthritis was noted as aninflammation and swelling of the paws, and diseasescores were assigned in a blinded analysis. The

Fig 5. Arthritis development. Arthritis development was monitored for mice of different strains in the pristane

model. (A) Severity and penetrance of arthritis measured by clinical scoring in mice of different strains 7 months

after pristane injection. Results are a composite of 3 studies. Mice at 7 months after pristane injection were sub-

jected to luminol bioluminescent imaging for detection of inflammation. (B, C) Images and quantitation of BALB/

c mice. (D, E) Results for DBA/1 mice. Control micewere treated with phosphate-buffered saline (PBS) instead of

pristane. Imaging results are a study representative of 2 experiments. ND, no data; Pris, pristane.

Translational ResearchVolume 163, Number 6 Bender et al 525

penetrance and severity was greater in the DBA/1 micecompared with the BALB/c mice (Fig 5, A), and thearthritis typically manifested earlier (data not shown).The arthritis was usually most robust in the ankles.Although clinical scoring by a blinded observer is an

effective means of measuring arthritis in these studies, amore quantitative and less subjective detection methodwas also used. In vivo bioluminescent luminol imagingwas used as an arthritis quantitation method in theBALB/c and DBA/1 mouse strains after pristane injec-tion. Luminol is a substrate for MPO, which on meta-bolism emits light that can be measured in liveanimals using commercially available imaging sys-tems.12 This method is a real-time indicator of infil-trating, MPO-expressing phagocytes such asneutrophils,12 which have been implicated as diseasemediators in human rheumatoid arthritis.20 Thus, lumi-nol imaging can be used to measure the inflammation

associated with arthritis in mouse paws. We found thatboth DBA/1 and BALB/c mice had elevatedMPO activ-ity in their arthritic paws (Fig 5, B and C, and Fig 5, Dand E, respectively). The arthritis score was higher inDBA/1, whereas the luminol signal was highest inBALB/c. Although apparently inconsistent, these read-outs do measure different aspects of inflammation.However, they are consistent in being significantly upre-gulated in pristane-treated mice. Thus, the luminol im-aging technique may be useful for assessing a specificmechanistic aspect (phagocyte infiltration) of arthriticinflammation in these mice.

The IFN gene signature. One of the newer biomarkersin the lupus field is the IFN gene signature. ElevatedIFN-a has been associated highly with lupus, althoughreliable detection of this analyte at the protein levelhas proved challenging. However, an increase in expres-sion of IFN-regulated genes has also been noted in lupus

Translational Research526 Bender et al June 2014

patient whole blood and has been shown to correlatewith disease severity and certain autoantibodyspecificities.21,22 The identity of these IFN-regulatedgenes has been reported by several groups and,collectively, is termed the IFN gene signature.However, there is, as yet, no standardized set of genesthat constitute the IFN gene signature for diagnosticconsistency among researchers. The IFN genesignature was used recently as an effectivepharmacodynamic biomarker in clinical trials ofneutralizing antibodies against IFN-a23,24 and holdspromise as a marker for future clinical testing of othertherapies.We tested the mouse models to determine which

exhibit an IFN gene signature similar to that seen in pa-tients with lupus. In preliminary studies, we performedgene expression profiling of NZB/W mice peripheralblood at 2 months and 9 months of age using DNA mi-croarrays and, although pathway analysis showed upre-gulation of canonical pathways, including B-celldevelopment, B-cell signaling, and SLE- and otherautoimmune-related signaling, we failed to detect anelevation in IFN-regulated genes (data not shown). Ithas been reported that several IFN genes are increasedin pristane-treated mice and that IFN-I receptor(IFNAR) deficient mice are disease resistant in thismodel.25,26 Preliminary work indicated varyingdegrees of increased expression of IFN-regulated genesin the 3 strains treated with pristane, with BALB/cshowing a low degree of upregulation (data not shown).To expand on these findings, we evaluated the pristaneDBA/1 and SJL models using a broad IFN gene signa-ture qPCR TLDA panel. We found that the DBA/1mice expressed a much more robust IFN gene signaturein peripheral blood compared with the SJL mice (Fig 6,A). Of the 20 genes that had detectable expression in theDBA/1 mice, 14 of them were elevated statistically andsignificantly with disease. To quantify the results, anIFN score was calculated for each mouse (Fig 6, B;see Materials and Methods for a complete descriptionof the calculation method). There was a statistically sig-nificant difference in IFN score when comparingpristane-treated with PBS-treated mice in the DBA/1strain, but not in the SJL strain (Fig 6, B). This IFN scorecalculation can be used to quantify and compare effectsreadily of different treatments on the IFN gene signa-ture, as has been done in clinical trials.23,24

To examine the similarity of this murine disease-regulated gene set to human IFN-driven gene expres-sion, PBMCs were isolated from the blood of healthyvolunteers and treated with recombinant type I IFN iso-forms, and the expression of selected IFN-regulatedgenes was measured by qPCR using a custom humanTLDA. The concentration of IFN used was well above

physiological levels, to ensure clear signal. A heatmap of the results (Fig 6, C) shows that the PBMCIFN treatment increased greatly the expression of nearlyall the selected genes. Of the 20 genes with an expres-sion that was detectable in the pristane mice, 15 wereincluded on the human TLDA, and the expression levelsof 14 were regulated statistically by IFN treatment (onlyMMP9 was not). This result confirms that the pristanedisease-regulated genes in the mouse system are, infact, type I IFN regulated in humans, and verifies theutility of the qPCR TLDA approach.The IFN gene signature in lupus patient blood sam-

ples was evaluated for comparison with data derivedfrom the pristane disease mouse models to assesstranslatability. Blood samples were collected from 20healthy donors and 20 patients with lupus. Geneexpression levels were measured using DNA microar-ray analysis. The fold change in expression of selectedgenes in the patients with lupus relative to the healthycontrol subjects is shown by heat map in Fig 7, A.Sixteen of the 20 genes that had detectable expressionin mice (Fig 6, A) also showed detectable expression inthe human microarray analysis. All the genes elevatedin the DBA/1 pristane mice, except for ISG15, werealso increased significantly in the patients with lupus.In addition, LY6E and RSAD2 were elevated signifi-cantly in the human diseased samples whereas theywere not in the mice (Fig 8, A). Hierarchical clusteringwas performed and the patients with lupus clearlysegregated into two groups of either IFN low or IFNhigh based on their expression levels of IFN-regulated genes. IFN scores were calculated to quan-tify the results, and the 2 groups were found to bestatistically different, with a mean score of 0.87 forthe IFN-low group and a mean IFN score of 1.96 forthe IFN-high group (Fig 7, B). The IFN-high patientsalso showed upregulation of other systems, includinghematologic function and humoral response (data notshown), corresponding to functions summarized byArasappan et al.27 In addition, autoantibodies weremeasured by ELISA in plasma samples from the pa-tients. IFN-high patients also had higher levels of au-toantibodies for anti-dsDNA, anti-histones, and anti-RiboP (Figs 7, C–E) consistent with publishedresults.22 Of the 3 autoantibodies, anti-RiboP wasmost associated with the IFN gene signature; it hadthe best correlation with IFN score (Fig 8, B;Spearman correlation coefficient 5 0.656). Thisfinding in humans is analogous to the elevated RiboPreactivity and high IFN gene signature in the DBA/1pristane model because we also observed thatpristane-treated diseased mice had both high RiboP ti-ters and a high IFN score (Fig 8, C). However, themagnitude of the anti-RiboP titer and IFN score

Fig 6. Analysis of the interferon (IFN) gene signature in pristane mice and human peripheral blood mononuclear

cells (PBMCs). Blood samples were collected from phosphate-buffered saline (PBS)-treated healthy mice and

pristane-injected diseased mice of the DBA/1 and SJL strains. RNAwas isolated and quantitative polymer chain

reaction (qPCR) Taqman low-density array (TLDA) analysis was performed to quantitate the expression of IFN-

regulated genes. (A) Heat map indicating which of the genes were most highly induced with disease (red) in the 2

different models. An ‘‘IFN score’’ was calculated for eachmouse by determining the median fold change for all the

genes in the panel. (B) The values for each mouse are plotted. The results shown are representative of 2 SJL pris-

tane studies and 4 DBA/1 pristane studies. (C) Human PBMCs isolated from healthy donors were treated with a

cocktail of recombinant IFN (rIFN) subtypes, and the expression of IFN-regulated genes was measured using a

qPCR TLDA panel. The heat map of the results shows the increased expression of genes after 20 hours of IFN

treatment. Changes in expression are shown relative to vehicle-treated cells at 3 hours. PBMCs from 3 healthy

donors were tested. *Significantly different from PBS (P , 0.01). Ctl, control; Pris, pristane.

Translational ResearchVolume 163, Number 6 Bender et al 527

were not well correlated (Spearman correlationcoefficient 5 0.045 for DBA/1 pristane mice).

DISCUSSION

Studies conducted in mouse lupus models havegreatly informed the field about aspects of human dis-ease etiology and have promoted drug discovery.5,9,28

However, exploratory drug activity in preclinicalanimal studies has not translated consistently into

successful development of effective lupus therapies, asevidenced by the remaining high, unmet medical needin this disease area. This is the result, at least in part,of the extreme heterogeneity in the human conditionand of the disconnect between the disease seen in miceand humans. Mouse models have translated reasonablywell in the era of general immunosuppressivetherapies, but as the era of genomics gives rise to morefocused therapies, an understanding of the underlyingmechanisms of autoimmunity and end organ damage

Fig 7. Analysis of the interferon (IFN) gene signature and autoantibodies in patients with system lupus erythema-

tosus (SLE). Blood was collected from 20 patients with SLE and 20 healthy donors into heparinized tubes or

directly into PAXgene RNA tubes. PAXgene samples were subjected to DNA microarray analysis as described

in Methods. (A) Gene expression was determined and a fold change for a panel of IFN-regulated genes for the

patients with SLE was determined relative to the expression of the healthy donors. The heat map shows the

fold changes for the patients with SLE and the individuals are classified as ‘‘IFN low’’ or ‘‘IFN high,’’ according

to Manhattan hierarchical clustering based on their expression of the IFN-regulated genes. (B) An IFN score was

calculated for each patient and is the median fold change for all the genes shown in the heat map. (C–E) From the

whole blood samples, plasma was isolated and the titers of anti-dsDNA (C), anti-antihistones (D), and anti-

Ribosomal phosphoprotein P (RiboP) (E) were determined by enzyme-linked immunosorbent assay. Medians

are shown. *Significantly different (P , 0.01).

Translational Research528 Bender et al June 2014

in our models will be needed. One drug evaluationstrategy is to explore a drug target or candidatetherapy comprehensively in several different animalmodels that recapitulate different lupus patient subsetsto try and identify which patients with a specificdisease etiology or symptoms will benefit most fromthe drug. A more focused strategy is to select a singlemodel for drug evaluation based on the model’srelevance to the drug’s mechanism of action and itssimilarity to a specific target lupus patient population.To use either strategy, the underlying etiology anddisease manifestations of the mouse lupus models

must first be determined. The results presented herehelp define clinically relevant disease characteristicsof mouse lupus models that increase their applicabilityto basic lupus research, as well as drug discovery anddevelopment. We have summarized the findings forthe different models and their relevance in Table II.In characterizing the time course for disease develop-

ment, we noted significant differences in mouse lupusmodels. Pristane-treated mice developed anti-dsDNAautoantibodies as early as 3 months after pristane injec-tion, and levels had peaked by 5 months post-pristane(Fig 1). In contrast, the NZB/Wmice anti-dsDNA levels

Fig 8. Comparison of the interferon (IFN) gene signature in lupus-diseasedmice and humans. (A) AVenn diagram

was constructed to show the IFN-regulated genes in common between pristane-diseased DBA/1 mice and human

patients with lupus. For the 2 groups, genes that were found to be significantly increased with disease (mice in Fig

6,A, and humans in Fig 7,A) and were also elevated in human peripheral bloodmononuclear cells treated with IFN

(Fig 6, C) are shown on the diagram. The diseased mice and humans had 8 genes in common whereas 1 gene was

unique to mice and 2 to humans. (B, C) Plots were constructed and linear regression was performed to determine

the correlation of IFN score and anti-Ribosomal phosphoprotein P (RiboP) titers for human patients with lupus (B)

and DBA/1 pristane mice (C). Data shown for the patients with lupus was derived from Fig 7, whereas data for the

DBA/1 mice is a combination of 2 studies.

Translational ResearchVolume 163, Number 6 Bender et al 529

were elevated at 5 months of age and a peak was notreached until 8–9 months of age. Similarly, mortalityfor the SJL mice was noted as early as 3 months afterpristane treatment whereas NZB/W mice deaths werenot seen until 8 months of age (Fig 3). This confirmedthe time course for proteinuria development, whichwas earlier in the SJL mice (Fig 3). Accelerated diseasedevelopment in the pristane model is one considerableadvantage for this model over the NZB/W model.Anti-dsDNA autoantibodies are a hallmark of lupus

and are found almost universally in patients.30 However,many other autoantibody reactivities are seen in lupuspatient serum and may define patient subsets withdiffering disease course or organ system involvement.The heterogeneity in the reactivities found in patientswith lupus is substantial, and new technologies havebeen developed to provide a means for characterizingautoantibody profiles broadly.31 Results have shownthat lupus patient subsets can be defined by the clusters

of autoantibodies they exhibit,10,32 which in turn can belinked with underlying genetic variants known toassociate with disease risk.33 Also, earlier work has pro-filed the NZB/W model in comparison with MRL-lprmice and BXSB mice and has found differences be-tween the models and possible correlations to humansubsets as well.34 We used microarray profiling technol-ogy to define the reactivities expressed in different lupusmouse models and found that specific clusters differen-tiate the NZB/W and pristane models (Fig 2). Pristane-induced disease mice demonstrated high levels ofRNA-associated autoantibodies (U1-snRNP, RiboP),which are nearly completely absent in this model on ge-netic deletion of the IFNAR2 chain of the type I IFNreceptor35 or the Toll-like receptor 7 receptor.36

Conversely, NZB/W mice had little to no detectableRNA-associated reactivity, but showed elevated levelsof autoantibodies against nuclear antigens such as chro-matin and histone proteins, which have been linked

Table II. Mouse lupus model characteristics

Model Disease symptoms Disease time courseCharacteristicautoantibodies

Whole-blood IFNgene signature

NZB/W Nephritis, mortality Slow Anti-chromatinAnti-histones

Absent

Pristane-Balb/c Arthritis, nephritis Slow Anti-U1-snRNPAnti-RiboP

Weak

Pristane-DBA/1 Arthritis Moderate Anti-U1-snRNPAnti-RiboP

Strong

Pristane-SJL Nephritis, mortality Rapid Anti-RiboP Moderate

Abbreviations: IFN, interferon; NZB/W, New Zealand black/white; U1-snRNP, U1 small nuclear ribonucleoprotein; RiboP, Ribosomal phosphopro-tein P.

Translational Research530 Bender et al June 2014

mechanistically to nephritis in this strain.37 The differ-ences observed between the mouse models were strik-ing and bear some resemblance to differences seen inhuman profiles.10,22,32 Profiling of patients with lupusalso revealed patient subsets defined by high levels ofantihistone reactivities or anti-U1-snRNP reactivities.10

The NZB/Wand pristane models may reflect the diseaseetiology within these different patient subsets and thusoffer better translational value to predict clinical successwithin an appropriately stratified patient subset. Whatdifferences in disease etiology leads to these distinc-tions in autoantibody reactivities is unknown and maybe explored using the different mouse models.In the pristane-induced lupus model, major differ-

ences were found between the different strains in theirend organ characteristics of disease. It is unclear whatcauses the differences in disease among these strains,but it is known that they differ in their immunologicbackground. Satoh et al8 have demonstrated that MHChaplotype as well as non-H2-associated genes can influ-ence autoantibody specificity and magnitude as well asend organmanifestation, and similar results are reportedin the NZB/W model. In the pristane model, forinstance, anti-RiboP autoantibodies are produced byall H2s and H2b strains tested, but are not producedby H2d or H2q mice,8 whereas pristane-inducedarthritis is observed only in BALB/c (H2d) and DBA/1 (H2q) strains and not in others such as B6 and B10(H2b) or SJL/J (H2s).38 Differences in disease subsetsin human patients with lupus are also recognized.39,40

For example, lupus nephritis occurs in a subset ofpatients whereas arthritis also develops in a subset ofpatients, but the latter is noted more widely. Usingeither SJL or DBA/1 pristane mice may allow forstudying targets or drugs intended to treat specificlupus patient subsets. The BALB/c pristane miceshowed some characteristics of both nephritis andarthritis, but with lower penetrance, severity, andslower time to development compared with the SJLand DBA/1 mice, making the BALB/c pristane modelless useful. Elucidation of the molecular pathology

that underlies these differences in target organpathology might have implications for understandinghuman lupus disease.IFN has gained increasing acceptance as a contrib-

uting factor to lupus development and pathogenesis,and several anti-IFN antibodies are currently in clinicaltrials for lupus treatment (sifalimumab and rontalizu-mab). High IFN activity is detected in many patientswith lupus, and associations have been made betweendisease severity or specific autoantibody levels andhigh IFN activity.41,42 However, because directdetection of IFN-a itself is problematic, theexpression level of a collection of genes regulated byIFN, termed the IFN gene signature,43 has been recog-nized as a surrogate for IFN-a and has been used as abiomarker in lupus clinical trials.23 However, the IFNgene signature has not been studied widely in lupusmouse models, and to our knowledge the current studyis the first to compare directly the broader IFN signaturein mouse pristane-induced lupus disease model with hu-man lupus samples. We did not find increased expres-sion of IFN-regulated genes in NZB/W gene profilingexperiments (data not shown), but did detect a robustIFN gene signature in pristane mice (Fig 6), consistentwith previous reports on IFN in the pristane model.25,26

We found that DBA/1 pristane mice had a particularlyrobust IFN gene signature, and this model offers ameans to evaluate a potential drug’s ability to affectthis important biomarker. The robust IFN genesignature is unique to this model; to our knowledge, ithas not been described in the NZB/W, MRL-lpr, orBXSB-Yaa models. Upregulation of Mx1 and PRKRhas been described in NZM2328 mice,44 but in anotherreport on B6.NZM congenics, the signature requiredcell sorting to be detectable.45 The ease of detectionof the signature in whole blood of the DBA/1 pristanemodel may make it particularly valuable for transla-tional purposes. We also found that most of the genesthat comprised our panel were regulated strongly byIFN treatment in human PBMCs, although it is impor-tant to note that the level of IFN used was greater than

Translational ResearchVolume 163, Number 6 Bender et al 531

expected in patients with lupus. Many of the same geneswere also increased in human patients with SLE (Fig 8,A), and a high IFN gene signature correlated with higherautoantibody titers, particularly anti-RiboP (Fig 8, B).This correlation of IFN activity with anti-RiboP titersin patients has been noted before,22 and it is interestingthat DBA/1 mice, which displayed a robust IFN genesignature in our studies, also demonstrated exceptionallyhigh RiboP titers (Fig 8,C). Based on these results, thereis a potential translation of the DBA/1 pristane mousemodel to high IFN gene signature patients with lupus.Although translating preclinical animal results to hu-

man disease can be problematic, there are clinicallyrelevant studies that can be run in mouse lupus models.Understanding the pathogenesis and characteristics ofthe animal models and how they recapitulate specific as-pects of human disease is imperative to choosing and us-ing them successfully. Matching models correctly topreclinical needs can allow us to define pharmacody-namic markers, understand the involvement of a drugtarget in disease pathogenesis, or choose which bio-markers to measure in clinical trials of the therapy orwhich patient subsets may benefit most from treatment.Our results illustrate and define the characteristics,strengths, and weaknesses of several mouse lupusmodels and their relationship to human disease. In ourstudies, we found the pristane models of the DBA/1and SJL strains to be particularly valuable, and theyoffer advantages over the NZB/Wmodel for basic lupusresearch as well as drug discovery and developmentbecause of their practicality and manifestation of clini-cally relevant features. The findings of this report willhelp to increase the value of preclinical studies for lupusdrug discovery and development by providing a betterunderstanding of various mouse lupus models and theirtranslatability to the human disease.

ACKNOWLEDGMENTS

Conflicts of interest: All authors have read the jour-nal’s policy on conflicts of interest. All authors wereemployees of Eisai Inc, at the time that the work re-ported in this manuscript was performed.The authors gratefully acknowledge Eva Skokanova,

Laurette Burgess, Laureen Knowles, and Marc Jean-Baptiste for contributing to animal care and handling,and Dr Quan Li of the UT Southwestern MicroarrayCore for directing the antigen microarray analysis.

Supplementary data

Supplementary data related to this article can befound at http://dx.doi.org/10.1016/j.trsl.2014.01.003.

REFERENCES

1. Liu Z, Davidson A. Taming lupus: a new understanding of path-

ogenesis is leading to clinical advances. Nat Med 2012;18:

871–82.

2. Harley IT, Kaufman KM, Langefeld CD, Harley JB, Kelly JA. Ge-

netic susceptibility to SLE: new insights from fine mapping and

genome-wide association studies. Nat Rev Genet 2009;10:

285–90.

3. Almqvist N, Winkler TH, Martensson IL. Autoantibodies: focus

on anti-DNA antibodies. Self/Nonself 2011;2:11–8.

4. Lu R, Robertson JM, Bruner BF, et al. Multiple autoantibodies

display association with lymphopenia, proteinuria, and cellular

casts in a large, ethnically diverse SLE patient cohort. Autoim-

mune Dis 2012;2012:819634.

5. Perry D, Sang A, Yin Y, Zheng YY, Morel L. Murine models of

systemic lupus erythematosus. J Biomed Biotechnol 2011;2011:

271694.

6. Morel L, Wakeland EK. Susceptibility to lupus nephritis in the

NZB/W model system. Curr Opin Immunol 1998;10:718–25.

7. Satoh M, Kumar A, Kanwar YS, Reeves WH. Anti-nuclear anti-

body production and immune-complex glomerulonephritis in

BALB/c mice treated with pristane. Proc Natl Acad Sci U S A

1995;92:10934–8.

8. Satoh M, Richards HB, Shaheen VM, et al. Widespread

susceptibility among inbred mouse strains to the induction of

lupus autoantibodies by pristane. Clin Exp Immunol 2000;121:

399–405.

9. GelfandMC, Steinberg AD, Nagle R, Knepshield JH. Therapeutic

studies in NZB-W mice: I. Synergy of azathioprine, cyclophos-

phamide and methylprednisolone in combination. Arthritis

Rheum 1972;15:239–46.

10. Li QZ, Zhou J, Wandstrat AE, et al. Protein array autoantibody

profiles for insights into systemic lupus erythematosus and

incomplete lupus syndromes. Clin Exp Immunol 2007;147:

60–70.

11. Williams RO. Collagen-induced arthritis as a model for rheuma-

toid arthritis. Methods Mol Med 2004;98:207–16.

12. Gross S, Gammon ST, Moss BL, et al. Bioluminescence imaging

of myeloperoxidase activity in vivo. Nat Med 2009;15:455–61.

13. Yung S, Chan TM. Anti-DNA antibodies in the pathogenesis of

lupus nephritis: the emerging mechanisms. Autoimmun Rev

2008;7:317–21.

14. Reeves WH, Lee PY, Weinstein JS, Satoh M, Lu L. Induction of

autoimmunity by pristane and other naturally occurring hydrocar-

bons. Trends Immunol 2009;30:455–64.

15. Yoshida H, Satoh M, Behney KM, et al. Effect of an exogenous

trigger on the pathogenesis of lupus in (NZB 3 NZW)F1 mice.

Arthritis Rheum 2002;46:2235–44.

16. Molino C, Fabbian F, Longhini C. Clinical approach to lupus

nephritis: recent advances. Eur J Intern Med 2009;20:447–53.

17. Weening JJ, D’Agati VD, Schwartz MM, et al. The classification

of glomerulonephritis in systemic lupus erythematosus revisited. J

Am Soc Nephrol 2004;15:241–50.

18. Xie C, Zhou XJ, Liu X, Mohan C. Enhanced susceptibility to end-

organ disease in the lupus-facilitating NZW mouse strain.

Arthritis Rheum 2003;48:1080–92.

19. Patten C, Bush K, Rioja I, et al. Characterization of pristane-

induced arthritis, a murine model of chronic disease: response

to antirheumatic agents, expression of joint cytokines, and immu-

nopathology. Arthritis Rheum 2004;50:3334–45.

20. Wright HL, Moots RJ, Bucknall RC, Edwards SW. Neutrophil

function in inflammation and inflammatory diseases. Rheuma-

tology 2010;49:1618–31.

Translational Research532 Bender et al June 2014

21. Karageorgas TP, Tseronis DD,Mavragani CP. Activation of type I

interferon pathway in systemic lupus erythematosus: association

with distinct clinical phenotypes. J Biomed Biotechnol 2011;

2011:273907.

22. Li QZ, Zhou J, Lian Y, et al. Interferon signature gene expres-

sion is correlated with autoantibody profiles in patients with

incomplete lupus syndromes. Clin Exp Immunol 2010;159:

281–91.

23. Yao Y, Higgs BW, Richman L, White B, Jallal B. Use of type I

interferon-inducible mRNAs as pharmacodynamic markers and

potential diagnostic markers in trials with sifalimumab, an anti-

IFNalpha antibody, in systemic lupus erythematosus. Arthritis

Res Ther 2010;12:S6.

24. McBride JM, Jiang J, Abbas AR, et al. Safety and pharmacody-

namics of rontalizumab in patients with systemic lupus erythema-

tosus: results of a phase I, placebo-controlled, double-blind,

dose-escalation study. Arthritis Rheum 2012;64:3666–76.

25. Lee PY, Kumagai Y, Li Y, et al. TLR7-dependent and FcgammaR-

independent production of type I interferon in experimental

mouse lupus. J Exp Med 2008;205:2995–3006.

26. Nacionales DC, Kelly-Scumpia KM, Lee PY, et al. Deficiency of

the type I interferon receptor protects mice from experimental

lupus. Arthritis Rheum 2007;56:3770–83.

27. Arasappan D, Tong W, Mummaneni P, Fang H, Amur S. Meta-

analysis of microarray data using a pathway-based approach

identifies a 37-gene expression signature for systemic lupus ery-

thematosus in human peripheral blood mononuclear cells. BMC

Med 2011;9:65.

28. Sang A, Yin Y, Zheng YY, Morel L. Animal models of molecular

pathology systemic lupus erythematosus. Prog Mol Biol Transl

Sci 2012;105:321–70.

29. Davidson A, Aranow C. Lupus nephritis: lessons from murine

models. Nat Rev Rheumatol 2010;6:13–20.

30. Isenberg DA, Manson JJ, Ehrenstein MR, Rahman A. Fifty years

of anti-dsDNA antibodies: are we approaching journey’s end?

Rheumatology 2007;46:1052–6.

31. Balboni I, Chan SM, Kattah M, Tenenbaum JD, Butte AJ, Utz PJ.

Multiplexed protein array platforms for analysis of autoimmune

diseases. Annu Rev Immunol 2006;24:391–418.

32. Ching KH, Burbelo PD, Tipton C, et al. Two major autoantibody

clusters in systemic lupus erythematosus. PLoS One 2012;7:

e32001.

33. Niewold TB, Kelly JA, Kariuki SN, et al. IRF5 haplotypes demon-

strate diverse serological associations which predict serum inter-

feron alpha activity and explain the majority of the genetic

association with systemic lupus erythematosus. Ann Rheum Dis

2012;71:463–8.

34. Li L, Nukala S, Du Y, et al. Murine lupus strains differentially

model unique facets of human lupus serology. Clin Exp Immunol

2012;168:178–85.

35. Thibault DL, Graham KL, Lee LY, Balboni I, Hertzog PJ, Utz PJ.

Type I interferon receptor controls B-cell expression of nucleic

acid-sensing Toll-like receptors and autoantibody production in

a murine model of lupus. Arthritis Res Ther 2009;11:R112.

36. Savarese E, Steinberg C, Pawar RD, et al. Requirement of Toll-

like receptor 7 for pristane-induced production of autoantibodies

and development of murine lupus nephritis. Arthritis Rheum

2008;58:1107–15.

37. Mortensen ES, Rekvig OP. Nephritogenic potential of anti-DNA

antibodies against necrotic nucleosomes. J Am Soc Nephrol

2009;20:696–704.

38. Wooley PH, Seibold JR, Whalen JD, Chapdelaine JM. Pristane-

induced arthritis: the immunologic and genetic features of an

experimental murine model of autoimmune disease. Arthritis

Rheum 1989;32:1022–30.

39. Petri M. Review of classification criteria for systemic lupus ery-

thematosus. Rheum Dis Clin North Am 2005;31:245–54.

40. Ruiz-Irastorza G, KhamashtaMA, Castellino G, Hughes GR. Sys-

temic lupus erythematosus. Lancet 2001;357:1027–32.

41. Dall’era MC, Cardarelli PM, Preston BT, Witte A, Davis JC Jr.

Type I interferon correlates with serological and clinical manifes-

tations of SLE. Ann Rheum Dis 2005;64:1692–7.

42. Weckerle CE, Franek BS, Kelly JA, et al. Network analysis of as-

sociations between serum interferon-alpha activity, autoanti-

bodies, and clinical features in systemic lupus erythematosus.

Arthritis Rheum 2011;63:1044–53.

43. Obermoser G, Pascual V. The interferon-alpha signature of sys-

temic lupus erythematosus. Lupus 2010;19:1012–9.

44. Sim DL, Bagavant H, Scindia YM, et al. Genetic complementa-

tion results in augmented autoantibody responses to lupus-

associated antigens. J Immunol 2009;183:3505–11.

45. Sriram U, Varghese L, Bennett HL, et al. Myeloid dendritic cells

from B6.NZM Sle1/Sle2/Sle3 lupus-prone mice express an IFN

signature that precedes disease onset. J Immunol 2012;189:80–91.