Content uploaded by L. Lightstone
Author content
All content in this area was uploaded by L. Lightstone on Dec 31, 2013
Content may be subject to copyright.
Urinary proteomic profiles distinguish between active and
inactive lupus nephritis
K. Mosley, F. W. K. Tam, R. J. Edwards
1
, J. Crozier, C. D. Pusey and L. Lightstone
Objectives. Key aims of the treatment of lupus nephritis (LN) are to induce and maintain remission with minimal side effects.
However, assessing ongoing renal inflammatory activity is poorly served by current diagnostic tests apart from renal biopsy,
but frequent biopsies cannot be justified. Our long-term aim is to identify novel biomarkers from urinary protein profiles to
improve diagnosis and monitoring of activity and response to therapy in LN.
Methods. We used surface enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) to
identify biomarkers able to discriminate between urine samples from patients with inactive (n¼49) and active (n¼26) LN.
Discriminant function analysis was used to define the minimum number of proteins whose levels best distinguished between
the two patient groups. Serial urines of six biopsied patients were studied prospectively, and multiple regression (MR) scores
calculated.
Results. Proteins with masses of 3340 and 3980 distinguished active from inactive LN with 92% sensitivity and specificity of
92% each. The prospective study of the biopsied patients demonstrated that MR scores could predict both relapse and remission
earlier than traditional clinical markers.
Conclusions. SELDI-TOF MS identified potential biomarker profiles strongly associated with activity in LN. Identification
of these proteins will allow us to devise specific assays to routinely monitor disease progression, and alter immunosuppressive
drug regimens accordingly. These proteins may also play a critical role in the pathogenesis of glomerulonephritis, and could
therefore provide targets for therapeutic intervention.
KEY WORDS: Lupus nephritis, Biomarkers, SELDI-TOF MS.
Introduction
Glomerulonephritis (GN) is a major cause of renal failure
and an important cause of morbidity and mortality in
multisystem diseases, such as systemic lupus erythematosus
(SLE) [1]. In SLE, active GN may be part of the initial
acute presentation, or may present at the time of relapse during
follow-up. Relapses are relatively frequent and early detection
is associated with better outcomes. Standard treatment involves
immunosuppressive therapy, which is associated with significant
side effects, including increased risk of sepsis and infertility.
Sensitive disease markers would help to optimize the initiation
and escalation of therapy at the time of active or relapsing disease,
and help to limit over-immunosuppression by improving detection
of significant renal remission.
Diagnosis of GN in SLE currently combines assessment
of renal function, quantification of proteinuria, and microscopy
and dipstick analysis of urine. These tests are useful in assessing
renal injury; however, they are poor indicators of ongoing
inflammatory activity. Serological markers, such as anti-double-
stranded DNA antibodies, are not specific for active renal
disease. Renal biopsies are important in the assessment of renal
disease. Examination of morphology and inflammatory cell
infiltration provides important information about the severity
of organ injury and disease activity. However, renal biopsy is
an invasive approach associated with significant risk, and
frequently repeated renal biopsies are not applicable in clinical
situations.
Two-dimensional gel electrophoresis (2-DE) has traditionally
been used to separate and analyse multiple proteins [2].
However, this technique is labour intensive, requires consider-
able technical expertise to produce reproducible gels and a
large amount of starting material is necessary. In addition, it
has only limited usefulness in detecting proteins of less than
20 kDa, hydrophobic proteins and those with extreme pIs.
An alternative to 2-DE-based methods are protocols based on
liquid chromatography-tandem mass spectrometry (LC-MS/MS),
where proteins are first digested with a suitable protease and
the peptides separated by LC. Tandem MS is then used to identify
the proteins from their fragmentation peptides [3]. This strategy
overcomes the problems associated with 2-DE, however it is usually
necessary to separate the proteins prior to LC-MS/MS to simplify
the mixtures, and relatively few samples can be run per day.
Surface enhanced laser desorption/ionization time-of-flight MS
(SELDI-TOF MS) is a relatively new approach that combines the
capture of proteins by biochemical or intermolecular interaction
with direct detection by MS [4]. Samples are applied to
ProteinChips with either chemically or biochemically modified
Renal Section and
1
Experimental Medicine and Toxicology, Division of Medicine, Imperial College London, Hammersmith Campus, London W12 0NN, UK.
Submitted 11 July 2006; revised version accepted 12 September 2006.
This work presented in part at the 38th Annual meeting of the American Society of Nephrology, Philadelphia, 2005. (Lightstone L, Mosley K,
Edwards RJ, Crozier J, Pusey CD and Tam FWK. Urinary proteomic profiles are able to discriminate between active and inactive lupus nephritis.
J Am Soc Nephrol 2005;16:8A.)
Correspondence to: K. Mosley, Renal Section, Division of Medicine, Imperial College London, Hammersmith Campus, London W12 0NN, UK.
E-mail: k.mosley@imperial.ac.uk.
Rheumatology 2006; 1 of 8 doi:10.1093/rheumatology/kel351
1of8
ß2006 The Author(s)
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted
non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Rheumatology Advance Access published November 14, 2006
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
surfaces to bind a subset of proteins. The bound proteins are then
ionized and desorbed, and the protein ions are separated by TOF
MS. Recent studies have demonstrated the usefulness of this
technique in identifying protein patterns in the urine associated
with renal allograft rejection [5–8], renal complications after
radiocontrast medium administration [9], urolithiasis [10] and
transitional cell carcinoma of the bladder [11]. The reproducibility
of this method for urine samples has been examined [12],
demonstrating that this is a promising platform for medium
throughput protein profiling. We have therefore set out to identify
protein profiles associated with lupus nephritis (LN) using
SELDI-TOF MS.
Patients and methods
Patient samples
Urine was collected from 26 active and 49 inactive SLE patients
during out-patient clinics. Urinary protein/creatinine ratios,
haematuria, anti-dsDNA antibody levels, serum creatinine and
complement C3 and C4 levels were all determined routinely. All
patients were previously confirmed as having renal involvement in
SLE by biopsy. At the time the sample was taken, all patients were
classified by the clinician, according to standard criteria, as having
active or inactive renal or systemic lupus. A diagnosis of active
renal disease depended on the presence of (i) a rise in creatinine
from baseline—a rise of >30 mol/l was considered significant if
the baseline was <150 mol/l or >50 mol/l if the baseline was
>150 mol/l and not attributable to other causes e.g. dehydration,
introduction of rennin–angiotensin blockade or intercurrent
illness; or (ii) a rise of total urinary protein:creatinine ratio of
>50 from a baseline of <100 or >100 from a baseline of >100, or
(iii) the new onset of haematuria [>10 red blood cells/high power
field (hpf)] by microscopic examination. If active renal disease was
suspected, a renal biopsy was performed where patients agreed.
In the absence of a renal biopsy, active SLE GN was only defined
if the criteria above were fulfilled and immunosuppression was
escalated. Patients were defined as in remission if their protein/
creatinine ratio was consistently <100 (equivalent to approxi-
mately 1g/24 h), urine microscopy showed <10 red blood cells/hpf,
and they had a stable serum creatinine, if in the normal range at
baseline, or a fall in serum creatinine by >50 mol/l if the baseline
was higher than 150 mol/l prior to change in immunosuppres-
sion. Serial samples were collected prospectively in several patients
before and after renal biopsy. Active systemic lupus was defined
on the basis of new onset of arthralgias, pericarditis or pleuritis,
myalgias, severe alopecia, typical rash or cytopenias not
attributable to drug treatments. Samples were also collected
from 15 normal controls. Midstream urine was collected and
stored at 48C overnight, as early optimization experiments
demonstrated that overnight storage did not affect the profiles
obtained. Samples were then centrifuged at 1400gfor 10 min to
remove cells that could release proteins after a freeze/thaw cycle,
and 1ml aliquots of the supernatant stored at 808C. Research
was carried out in accordance with the Declaration of Helsinki.
All patient and control samples were obtained with informed
consent and ethics approval by the Hammersmith, Queen
Charlotte’s and Chelsea and Acton Hospitals Research Ethics
Committee.
Protein profiling using surface enhanced laser
desorption/ionization time-of-flight mass
spectrometry (SELDI-TOF-MS)
Urine samples were thawed and centrifuged at 1400gfor 5 min.
Samples were diluted in the appropriate buffer to provide an equal
urinary creatinine concentration of 50 nmol/100 l.
To establish conditions for assessing urinary proteomics
to distinguish between disease groups, samples from normal
control, inactive and active LN patients were analysed using
normal phase (NP20), hydrophobic (H4), immobilized metal
affinity capture (IMAC30), strong anion exchange (SAX2, pH4
and pH9 buffers) and weak cation exchange (CM10, pH4 and
pH7) ProteinChips according to the manufacturer’s instructions
(Ciphergen, Freemont, CA, USA) with -cyano-4-hydroxycin-
namic acid (CHCA) and sinapinic acid (SPA) matrices (Ciphergen
and Fluka/Sigma-Aldrich, Poole, UK) and low and high laser
intensities (205 and 225). IMAC30 ProteinChips were found
to produce the most reproducible profiles, with numerous discrete
protein ions detected. SPA matrix at 100% saturation with
low laser intensity provided the best spectra in the range m/zrange
2500–100 000, and so these conditions were used for all further
experiments.
In subsequent experiments, IMAC-30 ProteinChips were
pre-treated with 100 mM CuSO
4
, neutralized with 0.1 M sodium
acetate and washed with binding buffer (0.1 M sodium
phosphate/0.5 M NaCl, pH 7). Either human serum albumin
(HSA 99% pure, fatty acid free, essentially globulin free,
Sigma-Aldrich) at concentrations of 1 g–100 g/spot (equivalent
to 100 g–10 mg/ml in urine) or LN urine samples were
then applied to the ProteinChip. Samples from patients with
urinary tract infections (UTIs) were excluded from the biomarker
discovery phase. In some experiments, albumin was spiked
into urine samples on the ProteinChip. Imidazole (0.1–10 mM,
Sigma-Aldrich) was also added to urine samples to compete
for binding to the ProteinChip surface. Quality control (QC)
samples, which consisted of pooled urine samples from two active
SLE, two inactive SLE and two control donors, were also applied
to the ProteinChips. Following sample application, ProteinChips
were incubated for 1h at room temperature with horizontal
shaking at 1050 rpm, washed three times in binding buffer
and rinsed in H
2
O. The spots were allowed to air dry for 10 min
before adding 2 0.5 l SPA matrix (Fluka). ProteinChips were
then read using the Ciphergen ProteinChip Reader IIc with
Ciphergen ProteinChip software, version 3.2. Data were collected
in the range 0–100 000m/zby averaging 82 laser shots using a laser
intensity of 205.
Data analysis
All samples were analysed in duplicate. The normalization factor
was calculated for each duplicate; all samples fell within the range
0.8–1.2. Mean intensity values were calculated for each protein
ion. To examine reproducibility, QC samples were assessed and
10 peaks were compared for each ProteinChip. The coefficient
of variation ranged from 17.0 to 25.5% (mean 22.4 0.6), which
is comparable with figures published previously [12, 13].
Initially, the peak clusters in the m/zrange 2500–100 000
were identified using the ProteinChip Biomarker Wizard software
version 3.2 for peaks with a signal-to-noise ratio of 2.5.
After correction for total protein, the data were normalized by
log-transformation. The identified protein ions that showed
significant differences between the groups (Student’s t-test) were
then examined visually to ensure that the peaks were discrete.
Any protein ions representing albumin (see subsequently) were
disregarded. The remaining protein ions were then analysed using
discriminant analysis (Systat, version 10.2, Richmond, CA, USA)
to identify combinations of proteins that best discriminate
between disease states. A logistic regression (LR) model was
also built using the same protein ions to calculate prediction
scores for each sample, allowing us to construct a receiver
operating characteristic (ROC) curve based on these values.
Specificity and sensitivity values for the traditional LN markers
urinary protein/creatinine ratio, serum creatinine, complement
factors three and four and double stranded DNA (dsDNA) were
also calculated for the patient groups.
2 of 8 K. Mosley et al.
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
For the prospective longitudinal study, multiple regression
(MR) scores were calculated using a linear equation derived from
the discriminant analysis canonical discriminant functions, where
Y¼aþ(b
1
X
1
)þ(b
2
X
2
)(a¼2.965, b
1
¼1.18, X
1
¼log10(m/z
3340 protein), b2¼0.853 and X2¼log10(m/z3980 protein).
An MR score of more than 0.4 was used to classify active
from inactive patients, this value being derived following
an examination of the MR scores of all of the correctly classified
patient samples as designated by the discriminant analysis.
Results
Patient characteristics
Patients and normal control donors were well-matched for
urinary creatinine content and age. Active SLE patients had
high levels of proteinuria and urinary protein/creatinine ratios,
with intermediate levels in inactive SLE samples and low levels
in normal controls. A large proportion of SLE patients were
female due to the sex-linked prevalence of this disease (Table 1).
Identification of urinary protein profiles associated
with SLE
Before analysing our full patient sample set, we investigated
the contribution of albumin to the protein profiles. Attempts
to remove albumin from the urine using immunoprecipitation
or a Vivapure anti-HSA kit (Vivascience, Hanover, Germany)
were unsuccessful for these samples (results not shown).
The reason for this was not established, although acceptable
depletion of serum albumin was achieved. To ensure that any
candidate biomarkers selected were not ions representing intact
albumin, we analysed HSA applied to IMAC30 ProteinChips
at concentrations ranging from 1 g to 100 g/spot (equivalent
to 100 g–10 mg/ml albuminuria), and examined the spectra
produced. In all, 21 protein ions were found that were derived
from the albumin preparation (Fig. 1). As the MS method used
cannot fragment ions the spectra produced would appear
to represent both singly (z¼1) and multiply (z>1) charged ions
of albumin and other proteins and peptides present in the albumin
preparation.
Urine samples from active and inactive LN patients were
analysed using IMAC30 ProteinChips, and the Ciphergen
Biomarker Wizard software was used to identify peaks represent-
ing protein ions with a signal-to-noise ratio of 2.5. In the m/z
range 2500–100 000, 80 protein ions were detected that were
present in more than 10% of the samples. After correction for
total protein, the data were normalized by log-transformation,
and those ions that showed significant differences in intensity
(P<0.01, Student’s t-test) between the two groups were selected
for further analysis. Non-discrete peaks were discounted (n¼36)
and peaks corresponding to albumin were also excluded from
further analyses.
To examine whether the amount of albumin present in the urine
could interfere with the expression of any potential biomarker,
we spiked active urine samples with purified albumin and
examined the effect on the six protein ions that showed the
greatest significant differences between groups (m/z3340, 3980,
4095, 4310, 5465 and 7965). There was no significant difference in
ion intensity on addition of up to 1mg/ml albumin (Fig. 2).
At 10 mg/ml, albumin reduced the expression of two of these
protein ions (m/z4095 and 4310) but had no effect on the remain-
ing four. In addition, albumin binding to the ProteinChip surface
FIG. 1. Protein ion profile of human serum albumin. Purified
human serum albumin was applied to IMAC30 ProteinChips
at an excess of 100 g/spot (equivalent to 10 mg/ml in urine) and
analysed using SELDI-TOF MS. (A) full m/zrange. (B) m/z
2500–5000. In all, 21 protein ion peaks with m/zof 2610, 2730,
2910, 3060, 3255, 3420, 3530, 3600, 3660, 3780, 4420, 8330, 9010,
13 785, 15 760, 22 150, 33 220, 44 420, 59 135, 66 370 and 88 405
were found that were derived from albumin.
FIG. 2. The effect of albumin on protein ion detection. Active
lupus nephritis urine samples were applied to IMAC30
ProteinChips either alone (unfilled bars) or in the presence of
0.1 (diagonal hatching), 1 (black fill) or 10 mg/ml human serum
albumin (stippled bars) and analysed using SELDI-TOF MS.
The six protein ions that previously showed the most significant
differences between samples from active and inactive patients were
examined. Results are expressed as fold change in relative ion
intensity of untreated urine. Asterisks denote significant difference
from control (P<0.01, Student’s t-test. n¼4).
TABLE 1. Patient characteristics
Active SLE
(n¼26)
Inactive SLE
(n¼49)
Normal control
(n¼15)
Proteinuria (g/l) 2.15 0.39 0.81 0.39 0.01 0.01
Urinary creatinine
(mmol/l)
9.79 0.98 10.72 0.87 8.2 2.1
Protein/creatinine
ratio (mg/mmol)
221.02 31.99 45.71 10.90 4.15 2.58
Age (yrs) 37.69 2.51 43.72 1.71 37.4 3.2
Sex, female (%) 84.6 90 50
Values are expressed as mean SEM.
Biomarkers of activity in lupus nephritis 3 of 8
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
was attenuated by addition of imidazole to the samples. In the
presence of 1mM imidazole albumin (m/z66370) and the protein
ions with m/zof 4095 and 4310 were significantly displaced
whereas the remaining protein ions remained bound to the
ProteinChip at this dose, suggesting that these proteins, which
have a greater affinity for the chip surface, were not related to
albumin (Fig. 3).
The 32 selected protein ions were analysed using discriminant
analysis. The protein ions that best discriminated between the
two groups were 3340 and 3980 resulting in the correct
identification of 92% (45/49) of the inactive and 92% (24/26)
of the active SLE patients (Table 2, Fig. 4). The addition of a
third protein ion with an m/zof 4095 only slightly improved
the classification (46/49 inactive, 24/26 active), and so further
analysis was performed on the 2-protein ion model. LR analysis
resulted in the selection of the same proteins as discriminant
analysis. An ROC curve constructed from the LR scores
gave a high area under the curve (AUC) value of 0.967
(Fig. 4). Comparison with the specificity and sensitivity
of traditional markers of LN calculated from this data set,
demonstrated that our markers had superior discriminatory
powers (Table 3).
Prospective longitudinal study
To confirm that our candidate biomarkers were responsive
to the disease state, we analysed serial samples from six patients
who had biopsy results indicative of active disease within 3 weeks
of urine collection (Fig. 5). Patient A’s renal biopsy was classified
as ISN/RPS class IV-G A/C, patient B’s as class IV-S A/C,
patient C’s as class III/IV A/C, patient D’s as class IV-S A/C,
patient E’s as class IV-G A/C and patient F’s as class III A.
The previous data established that an MR score of >0.4 denoted
active disease. At onset, the 4/6 defined as active clinically (A–D)
had appropriately high MR scores of between 0.877 and 2.37,
and the two clinically inactive patients (E–F) had lower scores
of 0.4 and 1.18. Both inactive patients relapsed during the
course of the study. Their MR score rose to active levels 3.5 and
2.7 months before clinical diagnosis of renal relapse. All four
active patients went into remission with a fall in the MR score
preceding clinical diagnosis of remission. Two of the six patients
(D and F) had multiple episodes of relapse and remission
that were reflected by parallel fluctuations in their score. Four
patients had samples collected during follow-up, which tested
positive for UTIs. In all cases, the presence of an infection did not
affect the predicted score.
Discussion
We have used a proteomic technique, SELDI-TOF MS,
to identify urinary protein ions associated with disease state in
SLE, and discovered candidate biomarkers that can distinguish
between active and inactive disease. Urine analysis provides
a unique method of sampling the local conditions of the whole
kidney. Distinctive urinary proteins may be generated not only
by increased renal synthesis, but also by a variety of factors
including a rise in circulatory levels, general effects of proteinuria
and haematuria, impaired tubular absorption of protein and
fragmentation of larger proteins [14–16]. A number of factors
need to be considered when using urine to detect biomarkers,
including variations in total protein content resulting from
proteinuria or hydration state, the presence of proteases and
variations in handling conditions, such as differences between
first-void and midstream urines. Studies by Schaub et al. [12] and
Rogers et al. [13] have examined some of these issues. They
demonstrated that it is necessary to use midstream urines,
but showed that the addition of protease inhibitors was found
to have little effect on profiles obtained, as did the time from
collection at 48C to freezing. We, therefore, incorporated these
findings into our protocol. Although SELDI-TOF MS is usually
performed on samples containing standardized protein levels,
in the case of proteinuric samples it would not be practical
to dilute these samples to the levels found in a normal control.
Preliminary studies demonstrated that proportionate protein
profiles patterns remain the same after dilution of the sample,
providing the less intense proteins are still detectable (data not
shown). We have therefore normalized according to creatinine
content, which removes concerns regarding the hydration levels
of the samples.
A major concern when examining proteinuric samples is the
albumin content, as albumin could potentially block most of the
binding sites on the ProteinChips, due to its molar excess,
or multiply charged albumin could be selected as a candidate
biomarker. Indeed, a recent paper [17] identified a protein ion
with an m/zof 66 kDa as a potential biomarker for acute renal
injury. In that study, although dipstick analysis of albuminuria
was negative their protein profiles clearly showed a range of peaks
corresponding to those that we observed with albumin alone.
We demonstrated that although 21 protein ion peaks were derived
from intact albumin, these were not amongst the protein ions we
selected as candidate biomarkers. Recent reports have described
the presence of large quantities of albumin fragments in normal
urine, particularly in the 300–500 Da range, derived from renal
degradation of protein [18–20]. The proportion of peptide to total
protein is reduced in renal disease and the fragments are larger
suggesting inhibition of albumin degradation [19, 21, 22].
We cannot rule out the possibility that our candidate biomarkers
FIG. 3. The effect of imidazole on protein ion detection.
To compete for binding to the ProteinChip surface, active lupus
nephritis urine samples were applied either alone (unfilled bars)
or in the presence of 0.1 (diagonal hatching), 1 (black fill)
or 10 mM imidazole (stippled bars). Protein spectra were detected
using SELDI-TOF MS. Albumin (m/z66370) and the six protein
ions that previously showed the most significant differences
between samples from active and inactive patients were analysed.
Results are expressed as fold change in relative ion intensity
of untreated urine. Asterisks denote significant difference between
control and 1 mM imidazole treatment (P<0.05, Student’s t-test.
n¼4). All protein ions showed a significant difference of
P<0.005 after treatment with 10 mM imidazole.
TABLE 2. Classification of inactive and active SLE urine samples
determined using protein ions with m/zof 3340 and 3980
Sample classified
Inactive SLE Active SLE Correctly identified (%)
Inactive (n¼49) 45 4 92
Active (n¼26) 2 24 92
4 of 8 K. Mosley et al.
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
are derived from albumin. If so, then they could represent either
specific cleavage products derived from differential expression of
proteases associated with disease activity or larger fragments
remaining after reduced degradation. Such fragments could
therefore still act as useful biomarkers as they would be specific
for activity. Recent studies [23, 24] demonstrated that albumin in
the serum, rather than being detrimental to SELDI-TOF MS
analysis, actually acts as a carrier protein, binding a majority
of low molecular weight markers (LMM) that would otherwise
be lost, resulting in amplification of the detectable levels of
markers. Here, urea/ 3-((3-Cholamidopropyl)dimethylammonio)-
1-propanesulfonate (CHAPS) denaturation of the urine samples
before application of the samples to the ProteinChips had no
effect on the profiles obtained (data not shown), suggesting that
matrix application and ionization sufficiently removes any LMM
from the albumin for detection. The lack of effect of the
denaturation step also demonstrates that LMM are not solely
retained on the ProteinChips due to their association with
albumin, suggesting that albumin does not saturate the binding
sites on the ProteinChip in our samples. Co-incubation of the
samples with imidazole also demonstrated that our candidate
FIG. 4. Candidate biomarker discovery. Urine samples from active (n¼26) and inactive (n¼49) lupus nephritis patients were analysed
using SELDI-TOF MS. Using discriminant analysis two protein ions with m/zof 3340 and 3980 were identified as best able
to differentiate between the groups. Panels (A) and (B) depict representative expression of candidate biomarkers in inactive patients and
active patients respectively. Asterisk denotes peaks derived from albumin, which were excluded from analysis. (C) Plot of log
10
normalized, total protein adjusted ion expression of the candidate biomarkers for active (filled circles) and inactive (open circles) samples.
RII denotes relative ion intensity. (D) ROC curve for the two biomarkers was generated using logistic regression scores, with an area
under the curve of 0.967.
TABLE 3. Comparison of traditional markers of lupus nephritis with
SELDI-TOF-MS derived biomarkers
Test Sensitivity (%) Specificity (%)
SELDI-TOF-MS 92 92
Urinary protein/creatinine 77 92
Haematuria 42 84
dsDNA 31 67
Serum creatinine 15 81
C4 13 89
C3 17 72
Biomarkers of activity in lupus nephritis 5 of 8
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
biomarkers are not bound to albumin, although interestingly, the
two protein ions that were affected by co-incubation with excess
albumin showed similar displacement patterns to albumin,
suggesting a possible association. Addition of imidazole to
samples in future analyses may further improve classification by
reducing albumin binding to the ProteinChip and allowing more
binding of less abundant proteins.
The main aim of this study was to detect proteins that are
closely associated with remission and relapse, which would allow
us to tailor the use of immunosuppressive drugs to each patient.
Analysis resulted in the correct identification of 92% of the
inactive patients (4/49 misclassified) and 92% of the active
patients (2/26 misclassified), with an ROC AUC of 0.967 using
protein ion peaks with m/zof 3340 and 3980. Although these two
protein ions provided the best separation, there were numerous
other strong candidates that could also be used. We found that
there was a high degree of correlation between the ion intensity of
the majority of protein ions, which limits the number of peaks
entered into the discriminant analysis model. These protein ions
could potentially be useful in devising future bench-based
methods. This early data is very promising, particularly as our
sample groups are relatively small. A pilot study performed
12 months earlier with a smaller number of patient samples
revealed consistent results (83% active and 86% inactive correctly
identified using M3340 and M3980; n¼18 and 22, respectively),
demonstrating the reproducibility of this technique with regard to
both the stability of the stored urine samples and the ProteinChip
reader setup. Comparison of the relative ion intensities of protein
ion peaks of patient samples present in both studies revealed no
significant change in expression of our candidate biomarkers
(M3340: active; study 1:25.2 7.3, study 2:25.9 4.8, P¼0.928.
Inactive; study 1:11.5 2.4, study 2:14.1 2.8, P¼0.481. M3980:
active; study 1:28.2 8.3, study 2:24.0 5.8, P¼0.684. Inactive;
study 1:10.12.4, study 2:9.52.0, P¼0.873), further confirming
reproducibility.
For the purpose of this study, we grouped all LN patients
together, regardless of their ISN/RPS biopsy classification.
Further work would be needed to determine if the approach
used here could be applied to individual classes of nephritis
to provide even greater accuracy of disease state assessment.
Post-analysis scrutiny of the ‘misclassified’ samples revealed
that three of the four samples from inactive patients were
previously classified by earlier biopsies as class V membranous
LN. They had negative serology and complement levels, but the
protein/creatinine ratios of 49, 106 and 116 were suggestive of
some persistent activity. The fourth misclassified inactive patient
FIG. 5. Sequential multiple regression scores for biopsied patients. Multiple regression scores were calculated for sequential urines
from six biopsied patients using the regression equation derived from the original analysis. Arrows depict time of biopsy. Filled circles
represent samples clinically classified as active and open circles those classified as inactive. The dashed line denotes a score of 0.4 above
which samples were classified as active by discriminant analysis of the original data set. Asterisks denote samples obtained while patients
had urinary tract infections.
6 of 8 K. Mosley et al.
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
has not been biopsied in the last 15 yrs, but had negative serology
and complement levels and a P/C ratio of 106 associated with
stable mild renal impairment. This patient is representative of one
of the most difficult clinical scenarios, namely deciding whether
mildly increased proteinuria in a long-term patient signifies
activity that warrants escalation of treatment. The two ‘mis-
classified’ active patients were also examined more closely. One of
these had normal anti-dsDNA and complement levels and a P/C
ratio of 122, suggesting that this may indeed be an inactive
patient. The remaining misclassified active patient had
abnormal serology and complement and a P/C of 103, however,
samples collected a month before and after the one used in this
study had normal P/C ratios, and the patient was classified
as inactive 7 months later. This may indicate that the patient
was improving at the time of analysis. These observations, rather
than highlighting problems with our test, demonstrate both
the weaknesses in current classification methods and the
potential of our candidate biomarkers to detect early signs of
relapse/remission.
To examine the responsiveness of our candidate biomarkers in
predicting activity and remission, we analysed serial samples from
six patients who had been biopsied within the course of our study.
In all cases, the MR score predicted a change in disease state prior
to clinical classification, suggesting that the MR score could detect
early onset of relapse and remission, thus allowing rapid tailoring
of treatment. There is a very real need to identify non-invasive
biomarkers associated with active LN for early diagnosis and
optimization of therapy, so that progression to end-stage renal
failure may be prevented. Numerous studies have attempted to
identify these, without much success [25]. However, recent
preliminary studies have identified potential urinary biomarkers
associated with renal activity and disease classification in SLE
[26–29]. Comparison of the specificity and sensitivity of our
candidate biomarkers with other non-invasive markers, and
the responsiveness of the biomarkers demonstrated during our
follow-up study suggests that we have developed a superior
method of determining activation state in LN.
Patients with UTIs were excluded from the initial biomarker
discovery experiments to prevent any bacterial proteins being
selected for further analysis. Of the six patients who were entered
for the prospective longitudinal study, four had UTIs on at least
one occasion. In all cases the infection had no effect on their
classification. This was particularly noteworthy for two inactive
samples, which retained their inactive classification. This further
validates our study, as false-positive classification due to infection
would have resulted in potential exclusion of all infected patients
during routine monitoring using our biomarkers. We cannot rule
out the possibility that these biomarkers are not specific for LN,
and aim to examine whether these protein peaks are present in the
urine samples of patients with non-LN. However, we did not set
out to identify markers that could classify patients as having LN,
but aimed to identify markers of activity, which may be useful
for the early diagnosis of renal relapse and monitoring of
immunotherapy.
The next stage in our study will be to analyse larger patient
groups and to run blinded samples to confirm the usefulness of
our currently identified biomarkers. After this confirmation,
we will then isolate and identify these biomarkers. Recent
studies have demonstrated that SELDI-TOF MS can be used to
provide reproducible results between sites [30, 31], and so
theoretically a diagnostic test could be based upon the candidate
biomarkers we describe here. However, our principle aim is
to use SELDI-TOF as a tool to detect proteins associated
with glomerular change in LN. Once these proteins have been
identified, we will devise assays, such as ELISA, to monitor renal
involvement in SLE.
In conclusion, we have identified two proteins in the urine of
patients with SLE using SELDI-TOF MS that allow us to
distinguish between patients in remission and those with active
renal disease without the need for invasive procedures.
Identification of these proteins will allow us to devise tests to
routinely monitor patients with LN, providing us with an
opportunity for early diagnosis of relapse and remission.
Therapy could therefore be tailored more accurately and promptly
to each patient, resulting in improved patient outcome.
Acknowledgements
This work was funded by the Hammersmith Hospital Trustee’s
Research Committee and the Anton Tardif Research Fund.
Funding to pay the Open Access publication charges for this
article was provided by Anton Tardif Research Fund.
The authors have declared no conflicts of interest.
References
1. Hill GS, Delahousse M, Nochy D et al. Outcome of relapse in lupus
nephritis: roles of reversal of renal fibrosis and response of
inflammation to therapy. Kidney Int 2002;61:2176–86.
2. O’Farrell PH. High resolution two-dimensional electrophoresis of
proteins. J Biol Chem 1975;250:4007–21.
3. Lin D, Tabb DL, Yates JRIII. Large-scale protein identification using
mass spectrometry. Biochim Biophys Acta 2003;1646:1–10.
4. Chapman K. The ProteinChip Biomarker System from Ciphergen
Biosystems: a novel proteomics platform for rapid biomarker
discovery and validation. Biochem Soc Trans 2002;30:82–7.
5. Schaub S, Rush D, Wilkins J et al. Proteomic-based detection of urine
proteins associated with acute renal allograft rejection. J Am Soc
Nephrol 2004;15:219–27.
6. Clarke W, Silverman BC, Zhang Z, Chan DW, Klein AS,
Molmenti EP. Characterization of renal allograft rejection by urinary
proteomic analysis. Ann Surg 2003;237:660–4.
7. O’Riordan E, Orlova TN, Mei JJ et al. Bioinformatic analysis of the
urine proteome of acute allograft rejection. J Am Soc Nephrol
2004;15:3240–8.
8. Schaub S, Wilkins JA, Antonovici M et al. Proteomic-based
identification of cleaved urinary beta2-microglobulin as a potential
marker for acute tubular injury in renal allografts. Am J Transplant
2005;5:729–38.
9. Hampel DJ, Sansome C, Sha M, Brodsky S, Lawson WE,
Goligorsky MS. Toward proteomics in uroscopy: urinary protein
profiles after radiocontrast medium administration. J Am Soc
Nephrol 2001;12:1026–35.
10. Cadieux PA, Beiko DT, Watterson JD et al. Surface-enhanced laser
desorption/ionization-time of flight-mass spectrometry (SELDI-TOF-
MS): a new proteomic urinary test for patients with urolithiasis.
J Clin Lab Anal 2004;18:170–5.
11. Vlahou A, Schellhammer PF, Mendrinos S et al. Development of a
novel proteomic approach for the detection of transitional cell
carcinoma of the bladder in urine. Am J Pathol 2001;158:1491–502.
Rheumatology
Key messages
Two candidate biomarkers have been
detected that can distinguish between
active and inactive lupus nephritis.
These biomarkers can predict early
relapse and remission.
Biomarkers of activity in lupus nephritis 7 of 8
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
12. Schaub S, Wilkins J, Weiler T, Sangster K, Rush D, Nickerson P. Urine
protein profiling with surface-enhanced laser-desorption/ionization
time-of-flight mass spectrometry. Kidney Int 2004;65:323–32.
13. Rogers MA, Clarke P, Noble J et al. Proteomic profiling
of urinary proteins in renal cancer by surface enhanced laser
desorption ionization and neural-network analysis: identification
of key issues affecting potential clinical utility. Cancer Res
2003;63:6971–83.
14. Cutillas PR, Chalkley RJ, Hansen KC et al. The urinary proteome
in Fanconi syndrome implies specificity in the reabsorption of
proteins by renal proximal tubule cells. Am J Physiol Renal Physiol
2004;287:F353–64.
15. Norden AG, Lapsley M, Lee PJ et al. Fragmentation of filtered
proteins and implications for glomerular protein sieving in Fanconi
syndrome. Kidney Int 2002;62:349.
16. Norden AG, Lapsley M, Lee PJ et al. Glomerular protein sieving
and implications for renal failure in Fanconi syndrome. Kidney Int
2001;60:1885–92.
17. Nguyen MT, Ross GF, Dent CL, Devarajan P. Early prediction
of acute renal injury using urinary proteomics. Am J Nephrol
2005;25:318–26.
18. Gudehithlu KP, Pegoraro AA, Dunea G, Arruda JA, Singh AK.
Degradation of albumin by the renal proximal tubule cells and the
subsequent fate of its fragments. Kidney Int 2004;65:2113–22.
19. Hilliard LM, Osicka TM, Clavant SP, Robinson PJ, Nikolic-
Paterson DJ, Comper WD. Characterization of the urinary albumin
degradation pathway in the isolated perfused rat kidney. J Lab Clin
Med 2006;147:36–44.
20. Strong KJ, Osicka TM, Comper WD. Urinary-peptide excretion by
patients with and volunteers without diabetes. J Lab Clin Med
2005;145:239–46.
21. Singh A, Gudehithlu KP, Le G et al. Decreased urinary peptide
excretion in patients with renal disease. Am J Kidney Dis
2004;44:1031–8.
22. Burne MJ, Panagiotopoulos S, Jerums G, Comper WD. Alterations
in renal degradation of albumin in early experimental diabetes
in the rat: a new factor in the mechanism of albuminuria. Clin Sci
1998;95:67–72.
23. Mehta AI, Ross S, Lowenthal MS et al. Biomarker amplification by
serum carrier protein binding. Dis Markers 2003;19:1–10.
24. Lowenthal MS, Mehta AI, Frogale K et al. Analysis of albumin-
associated peptides and proteins from ovarian cancer patients. Clin
Chem 2005;51:1933–45.
25. Illei GG, Tackey E, Lapteva L, Lipsky PE. Biomarkers in systemic
lupus erythematosus: II. Markers of disease activity. Arthritis Rheum
2004;50:2048–65.
26. Avihingsanon Y, Phumesin P, Benjachat T et al. Measurement of
urinary chemokine and growth factor messenger RNAs: a noninvasive
monitoring in lupus nephritis. Kidney Int 2006;69:747–53.
27. Rovin BH, Song H, Hebert LA et al. Plasma, urine, and renal
expression of adiponectin in human systemic lupus erythematosus.
Kidney Int 2005;68:1825–33.
28. Rovin BH, Song H, Birmingham DJ, Hebert LA, Yu CY,
Nagaraja HN. Urine chemokines as biomarkers of human systemic
lupus erythematosus activity. J Am Soc Nephrol 2005;16:467–73.
29. Oates JC, Varghese S, Bland AM et al. Prediction of urinary protein
markers in lupus nephritis. Kidney Int 2005;68:2588–92.
30. Rai AJ, Stemmer PM, Zhang Z et al. Analysis of Human Proteome
Organization Plasma Proteome Project (HUPO PPP) reference
specimens using surface enhanced laser desorption/ionization-time
of flight (SELDI-TOF) mass spectrometry: multi-institution correla-
tion of spectra and identification of biomarkers. Proteomics
2005;5:3467–74.
31. Semmes OJ, Feng Z, Adam BL et al. Evaluation of serum
protein profiling by surface-enhanced laser desorption/ionization
time-of-flight mass spectrometry for the detection of prostate
cancer: I. Assessment of platform reproducibility. Clin Chem
2005;51:102–12.
8 of 8 K. Mosley et al.
by guest on June 6, 2013http://rheumatology.oxfordjournals.org/Downloaded from
