Mapping similarities in mTOR pathway perturbations in mouse lupus nephritis models and human lupus nephritis.

Abstract

Treatment with sirolimus, a mammalian target of rapamycin (mTOR) inhibitor, has been shown to be efficacious in the MRL/lpr and NZB x NZW F1 mouse models of lupus nephritis, indicating a critical role for the mTOR pathway in both models. This type of demonstration of efficacy in animal models is usually a pre-requisite for advancement into clinical development. However, efficacy in an animal model often has not translated to the desired activity in the clinic. Therefore, a more profound understanding of the mechanistic similarities and differences between various animal models and human diseases is highly desirable.
Transcriptional profiling was performed on kidneys from mice with lupus nephritis; from mice who had efficacious drug treatment; and from mice before they developed nephritis. Analysis of variance with false discovery rate adjusted to p < 0.05 and an average fold change of two or more was used to identify transcripts significantly associated with disease and response to therapy. Pathway analyses (using various bioinformatics tools) were carried out to understand the basis for drug efficacy in the mouse model. The relevance in human lupus of the pathways identified in the mouse model was explored using information from several databases derived from the published literature.
We identified a set of nephritis-associated genes in mouse kidney. Expression of the majority of these returned to asymptomatic levels on sirolimus treatment, confirming the correlation between expression levels and symptoms of nephritis. Network analysis showed that many of these nephritis genes are known to interact with the mTOR pathway. This led us to ask what human diseases are linked to the mTOR pathway. We constructed the mTOR pathway interactome consisting of proteins that interact with members of the mTOR pathway and identified a strong association between mTOR pathway genes and genes reported in the literature as being involved in human lupus.
Our findings implicate the mTOR pathway as a critical contributor to human lupus. This broad pathway-based approach to understanding the similarities in, and differences between, animal models and human diseases may have broader utility.

Full text

Page 1

Open Access
Available online http://arthritis-research.com/content/10/6/R127
Page 1 of 16
(page number not for citation purposes)
Vol 10 No 6Research article
Mapping similarities in mTOR pathway perturbations in mouse
lupus nephritis models and human lupus nephritis
Padmalatha S Reddy1*, Holly M Legault1*, Joseph P Sypek2, Mark J Collins3, Elizabeth Goad4,
Samuel J Goldman2, Wei Liu1, Stuart Murray1, Andrew J Dorner5 and Margot O'Toole1
1Biological Technologies, Wyeth Research, Cambridge, 35 CambridgePark Drive, Massachusetts 02140, USA
2Inflammation, Wyeth Research, 200 CambridgePark Drive, Cambridge, MA 02140, USA
3Polysaccharides Quality, Wyeth Pharmaceutical, 1 Burtt Rd, Andover, MA 01810, USA
4Stryker Biotech, 35 South St, Hopkinton, MA 01748, USA
5ajDorner Consulting, 20 Baskin Road, Lexington, MA 02421, USA
* Contributed equally
Corresponding author: Margot O'Toole, motoole@wyeth.com
Received: 20 Jun 2008 Revisions requested: 15 Jul 2008 Revisions received: 22 Sep 2008 Accepted: 3 Nov 2008 Published: 3 Nov 2008
Arthritis Research & Therapy 2008, 10:R127 (doi:10.1186/ar2541)
This article is online at: http://arthritis-research.com/content/10/6/R127
© 2008 Reddy et al.; licensee BioMed Central Ltd.
This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Introduction Treatment with sirolimus, a mammalian target of
rapamycin (mTOR) inhibitor, has been shown to be efficacious
in the MRL/lpr and NZB × NZW F1 mouse models of lupus
nephritis, indicating a critical role for the mTOR pathway in both
models. This type of demonstration of efficacy in animal models
is usually a pre-requisite for advancement into clinical
development. However, efficacy in an animal model often has
not translated to the desired activity in the clinic. Therefore, a
more profound understanding of the mechanistic similarities and
differences between various animal models and human
diseases is highly desirable.
Methods Transcriptional profiling was performed on kidneys
from mice with lupus nephritis; from mice who had efficacious
drug treatment; and from mice before they developed nephritis.
Analysis of variance with false discovery rate adjusted to p <
0.05 and an average fold change of two or more was used to
identify transcripts significantly associated with disease and
response to therapy. Pathway analyses (using various
bioinformatics tools) were carried out to understand the basis
for drug efficacy in the mouse model. The relevance in human
lupus of the pathways identified in the mouse model was
explored using information from several databases derived from
the published literature.
Results We identified a set of nephritis-associated genes in
mouse kidney. Expression of the majority of these returned to
asymptomatic levels on sirolimus treatment, confirming the
correlation between expression levels and symptoms of
nephritis. Network analysis showed that many of these nephritis
genes are known to interact with the mTOR pathway. This led us
to ask what human diseases are linked to the mTOR pathway.
We constructed the mTOR pathway interactome consisting of
proteins that interact with members of the mTOR pathway and
identified a strong association between mTOR pathway genes
and genes reported in the literature as being involved in human
lupus.
Conclusions Our findings implicate the mTOR pathway as a
critical contributor to human lupus. This broad pathway-based
approach to understanding the similarities in, and differences
between, animal models and human diseases may have broader
utility.
Introduction
Clinical development of therapies is heavily dependent on
demonstrated efficacy in animal model(s), but efficacy in ani-
mal models often does not translate into clinical success. A
number of factors have been proposed as contributing to this
lack of concordance between efficacy in animal and clinical
studies [1-5]. One clear limitation of relying on disease models
in inbred strains is that the genes that produce the disease
phenotype in a given model may represent only a subset of the
genes that can cause the phenotype in complex human
ANOVA: analysis of variance; FDR: false discovery rate; H & E: haematoxylin & eosin; IFN: interferon; IPA: Ingenuity Pathway Analysis; IL: interleukin;
mTOR: mammalian target of rapamycin; PAS: periodic acid-Schiff; SD: standard deviation; SLE: systemic lupus erythromatous.

Page 2

Arthritis Research & Therapy Vol 10 No 6 Reddy et al.
Page 2 of 16
(page number not for citation purposes)
diseases such as lupus. Using our own animal model tran-
scriptomics, the vast and rapidly accumulating literature on
genes linked to human disease and pathway tools, we have
taken a broad analytical approach to identifying similarities
between the mouse and human 'lupus phenotype' at the level
of biological pathway perturbations. The potential advantage
of this approach is that, by linking the human disease pheno-
type to a pathway, drug development efforts can be targeted
to the pathway. Animal models with involvement of the same
pathway can then be chosen and/or derived.
Systemic lupus erythematosus (SLE) is a chronic inflammatory
autoimmune disease [6-8]. The pathophysiology of disease is
manifested by the production of autoantibodies directed
against multiple self-antigens. This dysregulation of the
immune system resulting in the loss of tolerance appears to be
mediated by both T cells and B cells. Many organs including
the kidney can be affected [9]. Direct action of autoantibodies,
deposition of immune complexes and pro-inflammatory
cytokines, particularly interferon (IFN) γ, have all been impli-
cated in disease pathophysiology [10-13].
There are at least four mouse models of lupus nephritis [14].
Both NZB × NZW F1 [15] and MRL/lpr mouse [16,17] strains
spontaneously develop autoimmune lupus nephritis. Female
mice from the NZB × NZW F1 cross (NZB/W) develop pro-
teinuria and only a small number (< 20%) survive to 52 weeks.
In MLR/lpr mice, the disease develops in both males and
females and is associated with the fas lpr mutation on the MLR
background. Mice develop significant proteinuria at 16 weeks
and show significant mortality rates (about 50%) by 20 weeks
[18,19]. Despite their independent derivation, lupus nephritis
in both MLR/lpr and NZB/W mouse models shows a remark-
ably efficacious response to sirolimus treatment [20-22].
Sirolimus (rapamycin) is an immunosuppressive drug that
binds to mTOR (mammalian target of rapamycin), a serine/
threonine kinase that regulates cellular proliferation and
metabolism and blocks G1 to S phase cell cycle progression,
interfering with T and B cell activation [23-25]. Sirolimus is
approved for the prevention of transplant rejection [26]. We
used our own data (presented here) and previously published
data on the efficacy of mTOR inhibitors in two mouse models
of lupus nephritis to infer that perturbations of the mTOR path-
way are critical to the development of lupus nephritis in both
these models. In order to assess the likelihood of mTOR path-
way involvement in human lupus, we examined the concord-
ance between the mTOR pathway interactome and genes
linked to human lupus and report the results of this analysis
here.
Materials and methods
NZB/W mice
NZB/W (NZB × NZW F1 cross) females were purchased
from the Jackson Laboratory (Bar Harbor, ME). These mice
were maintained and studied under pathogen free conditions
in accordance with guidelines from the American Association
for the Accreditation of Laboratory Animal Care and the Insti-
tutional Animal Care and Use Committee of Wyeth Research.
Experimental design
Beginning at 20 weeks of age, disease progression was mon-
itored weekly by assessing proteinuria. A cohort of mice,
selected at 20 weeks of age, served as the 'asymptomatic'
group (n = 4). Once fixed proteinuria of 30 to 100 mg/dL had
appeared on two consecutive occasions, (generally at 25 to
29 weeks of age), the diseased mice were randomly assigned
to either the sirolimus-treated group (n = 6 mice) or the
untreated (disease) group (n = 6). Sirolimus, dissolved in car-
boxymethylcellulose (vehicle), was subcutaneously adminis-
tered three times weekly in single doses of 1 mg/kg or 5 mg/
kg for eight weeks. Mice were monitored weekly until 52
weeks of age.
Assessment of proteinuria
Urine was manually expressed from each mouse on a weekly
basis, collected into a sterile container and assayed for the
presence of protein (specifically albumin) using a colorimetric
method (Albustix Reagent Strips, Bayer Corporation, Elkhart,
IN).
Proteinuria evaluations were scored as follows: grade 0.5 =
'trace' proteinuria; grade 1 = about 30 mg/dL; grade 2 =
about 100 mg/dL; grade 3 = about 300 mg/dL; and grade 4
= more than 2000 mg/dL. If mice achieved a grade 4 reading
on two consecutive days they were euthanised.
Assessment of renal pathology
Kidneys were harvested from mice one to four months after an
eight-week course of treatment. Three to five mice were exam-
ined in each group. One-half of a kidney was fixed by overnight
immersion in 10% formaldehyde and paraffin-embedded. The
other half was snap frozen for RNA preparation. To determine
the extent of renal damage, sections were stained with H & E
and periodic acid-Schiff (PAS) and scored for pathological
changes. In addition, glomerulopathy was scored on a 0 to 5
scale. Severity grades were as follows: 0 = normal or within
normal limits; 1 = minimal or slight; 2 = mild; 3 = moderate; 4
= marked; 5 = severe.
RNA purification and microarray hybridisation
Snap frozen murine kidney tissue was homogenised in RLT
buffer containing 1% beta-mercaptoethanol using the polytron
and RNA purified by Qiagen RNeasy columns (Qiagen, Valen-
cia, CA). Eluted RNA was quantified using a Spectramax 96-
well plate UV reader (Molecular Devices, Sunnyvale, CA, USA)
monitoring A260/280 OD values. The quality of each RNA
sample was assessed by capillary electrophoresis alongside
an RNA molecular weight ladder on the Agilent 2100 bioana-
lyser (Agilent Technologies, Palo Alto, CA, USA).

Page 3

Available online http://arthritis-research.com/content/10/6/R127
Page 3 of 16
(page number not for citation purposes)
Microarray processing
Five micrograms of total kidney RNA was prepared from indi-
vidual NZB/W F1 mice of the following groups; untreated
mice at 12 weeks (n = 4); untreated F1 mice at 36 and 42
weeks combined (n = 6); and sirolimus-treated mice at 36 and
42 weeks combined (n = 6). The animals selected for expres-
sion analysis reported here were representative of multiple
studies performed that confirmed the data on proteinuria, mor-
tality rates and histopathology here.
Biotin-labelled cRNA was prepared using an oligo T7-primed
reverse transcription reaction followed by in vitro transcription
reaction with biotin-labelled UTP and CTP. cRNA 15 μg was
fragmented and hybridised to Mu11KsubA and Mu11KsubB
arrays (Affymetrix, Santa Clara, CA). Hybridised arrays were
washed and stained with Streptavidin R-phycoerythrin (Molec-
ular Probes Inc., Eugene, OR) using the GeneChip Fluidics
Station 400 (Affymetrix, Santa Clara, CA) and scanned with a
Hewlett Packard GeneArray Scanner (Hewlett Packard Com-
pany, Palo Alto, CA) according to the manufacturer's proto-
cols. All array images were visually inspected for defects and
quality. Arrays with excessive background, low signal intensity
or major defects within the array were eliminated from further
analysis. GeneChip MAS 5.0 software was used to evaluate
the hybridisation intensity, compute the signal value for each
probe set and make an absent/present call. GeneChip signal
data for the samples analysed in this study are available under
Accession Number [GEO:GSE12924] [27].
Data normalisation and filtering
GeneChips were required to pass standardised quality control
criteria. RNA quality was monitored by the ratio of frequencies
measured by independent probe sets representing 5' and 3'
regions of glyceraldehyde 3-phosphate dehydrogenase. This
ratio must be more than 0.4. Filtering criteria for individual
probe sets required that a probe set was called 'present' or a
signal of 50 or more in at least one of the samples. All filtering
criteria were passed by 6384 probe sets and were subject to
the statistical analysis described below, and probe sets that
did not meet these criteria were not included in subsequent
analyses.
Hierarchical clustering
For hierarchical clustering of probe sets and arrays, the Log-2
scale MAS5 expression values from each probe set were first
z-normalised so each probe set had a mean expression level
of zero and a standard deviation (SD) of one across all sam-
ples. Then these normalised profiles were clustered hierarchi-
cally using an unweighted paired group method with
arithmetic mean, and the Euclidean distance measure.
Identification of genes associated with lupus nephritis
and response to sirolimus therapy
The disease-related fold change differences were calculated
by determining the difference in the log-2 signal of the 12-
week-old asymptomatic mice and the combined 36 and 42-
week-old diseased mice. Analysis of variance (ANOVA) was
performed using this metric to identify disease-related differ-
ences. Raw p values were adjusted for multiplicity of testing
using the false discovery rate (FDR) procedure of Reiner and
colleagues [28] using Spotfire (Somerville, MA). Genes with a
FDR p < 0.05 and an absolute fold change of two of more in
the comparison between disease and asymptomatic groups
were identified as being significantly associated with lupus
nephritis. Lupus nephritis genes were identified as being sig-
nificantly associated with response to sirolimus treatment if
they met an FDR p < 0.05, in comparison between sirolimus-
treated and disease groups. Those that failed to meet the FDR
p < 0.05 in the comparison of the sirolimus-treated group to
the disease group, did have a significant difference (FDR p <
0.05) in the comparison between sirolimus-treated and
asymptomatic groups, confirming a resistance to sirolimus
therapy.
Pathway analysis
Pathway analysis was performed using Ingenuity Pathways
Analysis (IPA) (Ingenuity Systems, Redwood City, CA), Meta-
Core (GeneGo Inc., St. Joseph, MI) and an in-house imple-
mentation of the sigPathway algorithm [29]. SigPathway is an
algorithm that identifies differentially expressed gene sets. An
FDR p < 0.01 was used to identify significantly changing gene
sets. Networks were built using genes that met FDR p < 0.05
within a gene set. Networks, canonical pathways and func-
tional processes for genes passing either the FDR p < 0.05
and an absolute fold change of two or more criteria and/or Sig-
Pathway filter FDR p < 0.05 were analysed using IPA. Human,
mouse and rat lupus-associated genes were identified using
the search tools within IPA and MetaCore. Rapalog-mTOR
pathway and its connectivity to the lupus nephritis disease
data and the lupus disease genes were explored using the
pathway building tools in IPA. The rapalog-mTOR pathway
interactome was built using IPA and all proteins were exported
to MetaCore to explore the human disease representation on
the mTOR pathway interactome. In MetaCore, a gene is con-
sidered to be associated with a condition (disease, pathologi-
cal or toxic process) as a biomarker if this gene or its product
(RNA or protein) has different properties in the disease and
healthy states. Such properties may include DNA characteris-
tics (eg, mutation, single nucleotide polymorphisms, chromo-
somal rearrangement); epigenetics (eg, methylation of a
promoter region, histone acethylation, nucleosome misplace-
ment); RNA level (eg, disease specific splice variant, higher/
lower expression level); or protein level (eg, disease-specific
isoform, mutant isoform, subcellular localisation, post-transla-
tional modification, abundance). These genes are used to gen-
erate human 'disease biomarkers' networks using direct
interactions between biomarker genes and proteins from their
MetaBase database. I2E (from Linguamatics Ltd., Cambridge,
UK) and clinical database findings were used to validate some

Page 4

Arthritis Research & Therapy Vol 10 No 6 Reddy et al.
Page 4 of 16
(page number not for citation purposes)
of the relationships between sirolimus analogues and various
human diseases identified in the mTOR pathway interactome.
Results
Short course therapy with sirolimus prevents onset of
murine lupus and renal damage
Treatment with sirolimus maintained 100% survival at age one
year, although survival in the control mice was only 20% (Fig-
ure 1a). Similarly, mice treated with sirolimus had minimal or no
increase in proteinuria and were asymptomatic for more than
three months after cessation of treatment (Figure 1b). Collec-
tively, these findings demonstrated the sustained benefit of a
short course of sirolimus therapy initiated early in disease.
To verify the therapeutic effects of sirolimus therapy on renal
pathology, kidney tissues were evaluated microscopically for
renal lesions and cellular infiltrates that were anticipated to
develop in NZB/W F1 mice at 36 weeks. Light microscopy of
kidney sections from vehicle-treated nephritic mice revealed
glomerulonephritis and interstitial inflammation, and also pro-
teinaceous tubular casts, consistent with their proteinuria (Fig-
ure 2b). Kidney sections from 12-week-old mice before
disease onset (Figure 2a) and from 36-week-old sirolimus-
treated mice (Figure 2c) revealed minimal renal pathology.
There was almost a complete absence of glomerular prolifera-
tion, interstitial infiltrates and casts. Histology scores for renal
inflammation, lymphocytic infiltrates and tubular atrophy are
shown in Table 1. There was good correlation between the
level of proteinuria and the severity of pathophysiological
changes observed in the kidneys. We have also collected
extensive data in this model showing a dramatic decrease in
anti-dsDNA titre with sirolimus treatment (see additional file 1).
Collectively, these findings confirmed the previously reported
beneficial effects of sirolimus treatment on the onset and
pathology of lupus nephritis in this mouse model [20,22,30].
Identification of the disease-associated transcriptome
RNA was prepared from tissue corresponding to one-half of a
kidney containing all cortex and medullary structures and har-
vested from asymptomatic mice at 12 weeks, diseased mice
at 36 and 42 weeks and sirolimus-treated mice at the same
age. Expression levels were assessed using Affymetrix Gene-
Chips (Affymetrix, Santa Clara, CA). There were 6384 probe
sets that met the criteria for inclusion in analysis. The expres-
sion patterns of these 6384 probe sets across groups were
visualised using an unsupervised clustering algorithm, which
assigns samples to clusters (nodes) based on similarity of
transcriptional pattern. A visual representation of differential
gene expression is shown in additional file 2. Samples were
grouped into three nodes: asymptomatic 12-week-old group;
36 and 42 week diseased group; and 36 and 42 week
sirolimus-treated group. This segregation indicated that the
renal RNA expression patterns of these three groups were dis-
tinct from each other. We then identified 1141 probe sets that
differed between the asymptomatic and 42-week diseased
groups with FDR p < 0.05 and an average fold change more
than 1.5. As seen in Figure 3, these 1141 probe sets showed
an almost identical change relative to the asymptomatic group
in the comparison with the 36-week diseased group (Pearson
correlation = 0.97).
Based on the similarities in the 36-week and 42-week mice,
ANOVA with FDR adjustment was performed comparing the
expression values of the disease group (consisting of
untreated 36- and 42-week-old mice) to those of the asympto-
matic 12-week-old group. This analysis yielded 195 differen-
tially expressed immunoglobulin probe sets and 547
differentially expressed non-immunoglobulin probe sets (using
significance criteria of FDR p ≤ 0.05 and a fold change differ-
ential ≥ 2). The disease-associated expression pattern of the
547 non-immunoglobulin transcripts included both up-regu-
lated and down-regulated non-immunoglobulin genes (Figure
4a). All 195 immunoglobulin probe sets were elevated in dis-
ease compared with asymptomatic animals (Figure 4b). Of
these 547 probe sets, protein interaction data from the litera-
ture is available in IPA for 387 genes. We have used this set
of 387 genes for pathway analyses as described below. The
complete list of non-immunoglobulin genes with functional
Figure 1
Proteinuria and survival measurements
Proteinuria and survival measurements. (a) Survival and (b) pro-
teinuria were measured weekly starting at 29 weeks of age (n = 10).
Grade 0.5 proteinuria = 'trace'; grade 1 = about 30 mg/dL; grade 2 =
about 100 mg/dL; grade 3 = about 300 mg/dL, a clinically significant
level; grade 4 = more than 2000 mg/dL; grade 5 = death.

Page 5

Available online http://arthritis-research.com/content/10/6/R127
Page 5 of 16
(page number not for citation purposes)
annotation is included in Additional file 3. An analysis of the
expression of these genes in kidneys of young (12 weeks old)
versus aged (32 weeks old) C57Bl/6 mice by ANOVA with
FDR adjustment showed no significant age-related changes in
the 547 transcripts associated with lupus nephritis.
Identification of nephritis-associated probe sets
modulated by sirolimus treatment
Of the 547 non-immunoglobulin probe sets associated with
nephritis at 36 and 42 weeks, 365 were modulated toward
asymptomatic levels by sirolimus treatment (with 150 that
were not modulated). Those that failed to meet the FDR p <
0.05 in the comparison of the sirolimus-treated group to the
disease group, did have a significant difference (FDR p <
0.05) in the comparison between sirolimus treated and asymp-
Table 1
Histopathology evaluation of kidneys from NZB/W F1 hybrid female mice at 42 weeks of age
Group AnimalInterstitial inflammationGlomerular nephritisTubular dilation
TherapyCellularityMembrane thickness
None1+2-2 +2+3
None2 +3 -1+3 +3
None3 +2 -1+3 +2
None4 +3-3 +3+4
None5+3 -3 +3+4
NoneMean+2.6 -2+2.8 +3.2
Sirolimus*1 (+1)0+1 +1
Sirolimus2 (+1)0 +10
Sirolimus3 (+2)000
Sirolimus40000
Sirolimus5(+1)000
SirolimusMean +10+0.4 0
p value, untreated versus treated0.0040.0020.00006 0.0001
0 = normal, 1 = slight, 2 = mild, 3 = moderate, 4 = marked, 5 = severe.
* denotes focal not diffuse.
Figure 2
Kidney histology. Representative sections of renal cortex from NZB × NZW F1 female mice
Kidney histology. Representative sections of renal cortex from NZB × NZW F1 female mice. (a) Section of kidney with glomeruli (arrows) and
tubules that are within normal limits from a 12-week-old mouse. (b) Section from an untreated 36-week-old mouse. Glomeruli (arrows) are small with
markedly thickened basement membranes and decreased numbers of cells (glomerulosclerosis). Tubules are dilated and contain protein casts or
droplets in the lumens. (c) Section from a 36-week-old mouse treated with sirolimus has kidneys that are within normal limits. Glomerular (arrows)
basement membranes are slightly thicker than those in the untreated 12-week-old mice. H & E, 40× original magnification.