Identification and Evaluation of Serum Protein Biomarkers That Differentiate Psoriatic Arthritis From Rheumatoid Arthritis

Abstract

Objective
To identify serum protein biomarkers that might distinguish patients with early inflammatory arthritis (IA) with psoriatic arthritis (PsA) from those with rheumatoid arthritis (RA) and may be used to support appropriate early intervention.

Methods
The serum proteome of patients with PsA and patients with RA was interrogated using nano–liquid chromatography mass spectrometry (nano‐LC‐MS/MS) (n = 64 patients), an aptamer‐based assay (SomaScan) targeting 1,129 proteins (n = 36 patients), and a multiplexed antibody assay (Luminex) for 48 proteins (n = 64 patients). Multiple reaction monitoring (MRM) assays were developed to evaluate the performance of putative markers using the discovery cohort (n = 60 patients) and subsequently an independent cohort of PsA and RA patients (n = 167).

Results
Multivariate machine learning analysis of the protein discovery data from the 3 platforms revealed that it was possible to differentiate PsA patients from RA patients with an area under the curve (AUC) of 0.94 for nano‐LC‐MS/MS, 0.69 for bead‐based immunoassay measurements, and 0.73 for aptamer‐based analysis. Subsequently, in the separate verification and evaluation studies, random forest models revealed that a subset of proteins measured by MRM could differentiate PsA and RA patients with AUCs of 0.79 and 0.85, respectively.

Conclusion
We present a serum protein biomarker panel that can separate patients with early‐onset IA with PsA from those with RA. With continued evaluation and refinement using additional and larger patient cohorts, including those with other arthropathies, we suggest that the panel identified here could contribute to improved clinical decision making.

Full text

81
Arthritis & Rheumatology
Vol. 74, No. 1, January 2022, pp 81–91
DOI 10.1002/art.41899
© 2021, American College of Rheumatology
Identication and Evaluation of Serum Protein
BiomarkersThat Dierentiate Psoriatic Arthritis From
Rheumatoid Arthritis
AngelaMc Ardle,1 AnnaKwasnik,1 AgnesSzentpetery,1 BelindaHernandez,2 AndrewParnell,1 Wilcode Jager,3
Sytzede Roock,4 OliverFitzGerald,1 and Stephen R.Pennington1
Objective. To identify serum protein biomarkers that might distinguish patients with early inammatory arthritis
(IA) with psoriatic arthritis (PsA) from those with rheumatoid arthritis (RA) and may be used to support appropriate
early intervention.
Methods. The serum proteome of patients with PsA and patients with RA was interrogated using nano– liquid
chromatography mass spectrometry (nano- LC- MS/MS) (n = 64 patients), an aptamer- based assay (SomaScan)
targeting 1,129 proteins (n = 36 patients), and a multiplexed antibody assay (Luminex) for 48 proteins (n = 64 patients).
Multiple reaction monitoring (MRM) assays were developed to evaluate the performance of putative markers using
the discovery cohort (n = 60 patients) and subsequently an independent cohort of PsA and RA patients (n = 167).
Results. Multivariate machine learning analysis of the protein discovery data from the 3 platforms revealed that it
was possible to differentiate PsA patients from RA patients with an area under the curve (AUC) of 0.94 for nano- LC-
MS/MS, 0.69 for bead- based immunoassay measurements, and 0.73 for aptamer- based analysis. Subsequently, in
the separate verication and evaluation studies, random forest models revealed that a subset of proteins measured
by MRM could differentiate PsA and RA patients with AUCs of 0.79 and 0.85, respectively.
Conclusion. We present a serum protein biomarker panel that can separate patients with early- onset IA with PsA
from those with RA. With continued evaluation and renement using additional and larger patient cohorts, including
those with other arthropathies, we suggest that the panel identied here could contribute to improved clinical decision
making.
INTRODUCTION
Psoriatic arthritis (PsA) is a form of inammatory arthritis
(IA) affecting ~0.25% of the population (1– 4). It is a highly het-
erogeneous disorder associated with joint damage, disability,
disguring skin disease, and poor patient- related quality of life
outcome measures (4). Inherently irreversible and frequently pro-
gressive, the process of joint damage begins at or before the clin-
ical onset of disease. Indeed, structural joint damage, which is
likely to result in joint deformity and disability, is present in 47%
of patients within 2 years of disease onset (3,5). Reductions in
quality of life and physical function are comparable to those in
rheumatoid arthritis (RA) and are compounded by the presence
of chronic disguring skin disease (6– 9). Direct and indirect health
costs pose a signicant economic burden on society and increase
with severe physical dysfunction (9).
Early diagnosis and management of PsA leads to better
long- term outcomes; however, with no diagnostic laboratory test
available, the diagnosis is often delayed or missed, and this has sig-
nicant consequences for individuals with PsA (10– 12). At disease
onset, PsA often resembles other forms of arthritis including RA.
Despite the clinical similarities between PsA and RA, their distinctive
Supported by the European Commission under the EU FP7 Programme
Monitoring Innate Immunity in Arthritis and Mucosal Inammation (MIAMI)
(project grant 305266) and the Health Research Board (grant HRA- HSR-
2015- 1284). The University College Dublin Conway Institute is supported by
the Higher Education Authority of Ireland.
1Angela Mc Ardle, PhD, Anna Kwasnik, PhD, Agnes Szentpetery, MD,
Andrew Parnell, PhD, Oliver FitzGerald, MBBCh, BAO, MRCPI, MRCP,
MD, Stephen R. Pennington, PhD: University College Dublin, Dublin,
Ireland; 2Belinda Hernandez, PhD: Trinity College Dublin and University
College Dublin, Dublin, Ireland; 3Wilco de Jager, PhD: Wilhelmina Children
Hospital and University Medical Centre Utrecht, Utrecht, The Netherlands;
4Sytze de Roock, PhD: University Medical Centre Utrecht, Utrecht, The
Netherlands.
Drs. FitzGerald and Pennington contributed equally to this work.
Dr. Pennington owns stock or stock options in Atturos. No other
disclosures relevant to this article were reported.
Address correspondence to Oliver FitzGerald, MBBCh, BAO, MRCPI,
MRCP, MD, or Stephen R. Pennington, PhD, University College Dublin, UCD
Conway Institute of Biomolecular and Biomedical Research, Dublin, Ireland.
Email: oliver.tzgerald@ucd.ie or Stephen.Pennington@ucd.ie.
Submitted for publication December 8, 2020; accepted in revised form
June 8, 2021.

Mc ARDLE ET AL
82
|
pathologic manifestations often require different treatments. For
example, drugs targeting the interleukin- 12 (IL- 12)/IL- 23 and IL- 17
pathways, which are highly effective in psoriasis and PsA, are inef-
fective in RA, while drugs targeting B cells such as rituximab are
effective in RA but have not been proven benecial in PsA (4,13).
PsA is most often diagnosed when a patient presents
with musculoskeletal inammation in the presence of psoriasis
and in the absence of rheumatoid factor (RF). However, a clear
diagnosis can be difcult, as up to 10% of PsA patients may have
RF or anti– citrullinated peptide antibody (ACPA), and joint involve-
ment may precede the development of skin or nail psoriasis in
15% of patients with PsA (14). The Classication of Psoriatic Arthri-
tis (CASPAR) Study Group criteria are accepted as having high
sensitivity (98.7%) and specicity (91.4%) in classifying patients
with longstanding PsA (15). CASPAR criteria show reduced sen-
sitivity in patients with early disease (87.4%), though specicity is
improved (99.1%) (16). CASPAR criteria are valid when including
patients in research studies or in clinical trials, but it is recognized
that they should not be used for diagnosis and are of little value
therefore in a primary care or dermatology setting where specialist
rheumatologic expertise is very often not readily available (4,17).
An effective clinical laboratory test is needed to improve diagnosis
and clinical decision making in PsA.
Ideally, a clinical laboratory test should be based on an eas-
ily accessible biologic sample such as blood (10), and we there-
fore set out to discover serum- based biomarkers that could
discriminate between patients with PsA and those with RA. With
advances in multiplexed technologies, it has become possible
to simultaneously mea sure multiple analytes. However, in com-
plex biologic uids such as serum, it is apparent that no single
technological platform is capable of measuring the entire protein
content of a given sample (3,4,18). For this reason, we under-
took a comprehensive and complementary analysis of the serum
proteome in a cohort of patients with early IA. We used unbiased
nano– liquid chromatography mass spectrometry (nano- LC- MS/
MS) for serum samples depleted of abundant proteins to iden-
tify differentially expressed proteins. In parallel, aptamer- based
and bead- based multiplexed assays were used to target low-
abundant proteins not easily detectable by nano- LC- MS/MS. Sta-
tistical analysis revealed that proteins identied by nano- LC- MS/
MS were the most useful in differentiating individuals with PsA
from those with RA. Therefore, in subsequent steps we prioritized
these proteins for further investigation.
The translation of biomarkers from discovery to clinical use
poses many challenges, not least because of the difculty of con-
dently identifying suitable candidates from the discovery phase.
Multiple reaction monitoring (MRM), a form of targeted MS, is a
highly versatile approach that makes it relatively easy to develop
and adapt assays that support the simultaneous mea surement
of multiple proteins. Assay development times are typically much
shorter for MRM assays compared to enzyme- linked immu-
nosorbent assays (ELISAs), and multiplexing of MRM assays
is signicantly easier. We therefore exploited the advantages of
MRM to undertake a 2- phase approach to progress the candi-
date protein biomarkers identied in the nano- LC- MS/MS discov-
ery study. First, we undertook a verication phase in which MRM
assays for a panel of 150 candidate biomarker proteins identied
in the discovery cohort were developed and used to measure
protein levels in patients from that cohort; in a second evaluation
phase, we adapted the MRM assay to encompass an expanded
panel of 173 proteins and used this to measure the proteins in an
independent cohort. Figure 1 provides an overview of the study
workow.
PATIENTS AND METHODS
Patients. In the discovery and initial verication phases, a
total of 64 patient samples were used, and the extensive clinical
characterization of the cohort has previously been described in
full by Szentpetery et al (19). Briey, patients ages 18 to 80 years
with recent- onset (symptom duration <12 months), treatment-
naive PsA or RA with active joint inammation were enrolled. PsA
patients (n = 32) fullled the CASPAR criteria (15), and patients with
RA (n = 32) met the American College of Rheumatology (ACR)/
European Alliance of Associations for Rheumatology (EULAR)
2010 classication criteria (20). Baseline serum samples were
obtained from each patient using standard methodology, aliquot-
ted, and frozen at −80°C (see Supplementary Document 1 on
the Arthritis & Rheumatology website at http://onlinelibrary.wiley.
com/doi/10.1002/art.41899/abstract). The study was approved
by St. Vincent’s Healthcare Group Ethics and Medical Research
Committee, and patients were enrolled only after agreeing to par-
ticipate in the study and providing informed consent.
Samples from a total of 167 patients were used in the second
verication phase. There were 95 patients recruited from a cross-
sectional cohort of patients with established PsA who all met
CASPAR criteria and 72 patients recruited from the RA Biologics
Registry of Ireland who all met ACR/EULAR 2010 classication
criteria and had similar levels of active disease as the PsA patients.
Again, baseline serum samples were obtained, aliquotted, and
frozen at −70°C.
Label- free nano- LC- MS/MS analysis. A detailed descrip-
tion of the unbiased LC- MS/MS workow has previously been
described (10). Briey, serum samples (1,700 μg) were depleted of
the 14 most abundant serum proteins (albumin, transferrin, hap-
toglobin, IgG, IgA, α1- antitrypsin, brinogen, β2- macroglobulin,
α1- acid glycoprotein, complement C3, IgM, apolipoprotein A- I,
apolipoprotein A- II, and transthyretin) using the Agilent Multiple
Afnity Removal System comprising a Hu- 14 column (HuMARS14)
(4.6 × 100 mm; Agilent Technologies, no. 5188- 6557) on a Biocad
Vision Workstation. Depleted fractions (containing 50 µg protein)
were reduced, denatured, and alkylated prior to trypsinization.
The digested samples were desalted and puried using C18 resin

BIOMARKERS DIFFERENTIATE PsA FROM RA
|
83
pipette stage tips. Puried samples were dried under vacuum
and resuspended in MS- compatible buffer A (3% acetonitrile,
0.1% formic acid) (21,22). Label- free nano- LC- MS/MS analysis
was performed on a Q Exactive mass spectrometer equipped
with a Dionex Ultimate 3000 (RSLCnano) chromatography sys-
tem (ThermoFisher Scientic). Two microliters (equivalent to 2 µg
of digested protein) of each sample were injected onto a fused
silica emitter separated by an increasing acetonitrile gradient over
101.5 minutes (ow rate 250 nl/minute) (10).
Bioinformatic data analysis. As previously reported,
nano- LC- MS/MS data were visually inspected using XCalibur
software (version 2.2 SP1.48). MaxQuant (version 1.4.12) was
then used for quantitative analysis of the LC- MS/MS data, while
Figure 1. Overview of the experimental workow. Three platforms were used: nano– liquid chromatography mass spectrometry (nano- LC- MS/MS),
aptamer- based immunoassays, and bead- based immunoassays for biomarker discovery. Resulting data were analyzed by univariate and multivariate
analysis. Putative biomarkers identied by nano- LC- MS/MS proteins were brought forward for multiple reaction monitoring (MRM) assay development,
which was divided into 2 phases. During phase I, it was possible to develop an assay for 150 proteins which were measured in the discovery cohort.
During phase II, an assay was developed for 173 proteins which were measured in an independent evaluation cohort. SA = streptavidin.

Mc ARDLE ET AL
84
|
Perseus software (version 1.5.0.9) supported statistical analysis
(10,23).
Aptamer- based analysis. Individual patient serum sam-
ples were subjected to a multiplexed aptamer- based assay devel-
oped by Gold et al to measure the levels of 1,129 proteins, as
previously reported (10).
Bead- based immunoassay. Individual serum samples
were subjected to in- house– developed and validated multiplexed
immunoassays measuring 48 analytes. The assays and analyses
were undertaken, as previously described, at the Multiplex Core
Facility Laboratory of Translational Immunology at the University
Medical Centre Utrecht (10).
MRM design and optimization. The development and
optimization of MRM assays was performed using Skyline soft-
ware (version 3.6.0.1062) (MacCoss Lab) (24). Assays for proto-
typic peptides were developed for all proteins of interest where
peptides showed no missed cleavages or “ragged ends” and
sequence length was between 7 and 25 amino acids. When
possible, peptide sequences with reactive cysteine or methionine
residues were avoided but not excluded. An MRM assay was
deemed to be analytically validated when it demonstrated the
following characteristics: dot product ≥0.8, signal to noise ≥10,
data points under the curve ≥10 (25), and percentage coefcient
of variance showing a retention time ≤1% and area ≤20% (26).
The majority of MRM assays developed signicantly exceeded
these criteria.
Sample preparation for LC- MRM analysis. Verication
phase. Crude serum (2 µl) was added to the wells of 96-well
deep well plates (ThermoFisher Scientic) and diluted at 1:50
with NH4CO3 (Sigma). RapiGest denaturant (Waters) was resus-
pended in 50 mM NH4CO3 to give a stock solution of 0.1%
weight/volume, and 50 μl of this stock solution was added to
each sample so that the nal concentration of RapiGest was
0.05%. Plates were covered with adhesive foil (ThermoFisher
Scientic), and samples were incubated in the dark at 80°C for
10 minutes. After incubation, plates were centrifuged at 2,000
relative centrifugal force (rcf) at 4°C for 2 minutes to condense
droplets. Subsequently, dithiothreitol (DTT) was added to each
sample at a nal concentration of 20 mM. Samples were then
incubated at 60°C for 1 hour followed by centrifugation at 2,000
rcf at 4°C for 2 minutes.
Next, iodoacetamide was added to each sample to give a
nal concentration of 10 mM, and plates were incubated at 37°C
in the dark for 30 minutes. Plates were again centrifuged at 2,000
rcf at 4°C for 2 minutes, and samples were then diluted with LC-
MS/MS– grade H2O to produce a nal concentration of 25 mM
NH4CO3. Trypsin (Promega) was added to each sample so that
the protein:enzyme ratio was 25:1. The reaction was stopped with
the addition of 2 μl of neat triuoroacetic acid(Sigma) to each
sample and incubated for a further 30 minutes at 37°C. In order
to pellet RapiGest, digests were transferred from 96- well plates to
1.5 ml low- bind Eppendorf tubes and centrifuged for 30 minutes
at 12,000 rcf. Supernatants were removed and transferred into
clean Eppendorf tubes and lyophilized by speed vacuum at 30°C
for 2 hours. Lyophilized samples were stored at −80°C until further
use.
Evaluation phase. The denaturant used previously (Rapi-
Gest) was substituted with 25 µl denaturant solution comprising
50% triuoroethanol in 50 mM NH4HCO3 with 10mM DTT, and
this mitigated the need for the high- speed spin and transfer of
supernatant, which represented an additional processing step
less compatible with 96- well plate workows.
MRM analysis. MRM analysis was performed using an
Agilent 6495A triple- quadrupole mass spectrometer with Jet-
Stream electrospray source (Agilent) coupled to a 1290 Qua-
ternary Pump HPLC system. Peptides were separated using
analytical Zorbax Eclipse Plus C18 (rapid resolution HT 2.1 ×
50 mm, 1.8um, 600- bar columns) (Agilent) before introduction
to the triple- quadrupole mass spectrometer. A linear gradient of
acetonitrile (99.9% acetonitrile, 0.1% formic acid) 3– 75% over
17 minutes was applied at a ow rate of 0.400 µl/minute with
a column oven temperature of 50°C. Source parameters were
as follows: gas temperature 150°C, gas ow 15 liters/minute,
nebulizer psi 30, sheath gas temp 200°C, and sheath gas ow
11 liters/minute. Peptide retention times and optimized collision
energies were supplied to MassHunter (B0.08; Agilent Technolo-
gies) to establish a dynamic MRM- scheduled method based on
input parameters of 800- msec cycle times and 2- minute reten-
tion time windows. The percentage coefcient of variation (%CV)
of biologic and technical replicates was used as a measure of
variance and was calculated using the following standard calcu-
lation: %CV = (SD/mean) × 100.
ELISA analysis. C- reactive protein (CRP) levels were eval-
uated at St. Vincent’s University Hospital using an automated
CRPL3 Tina- quant assay (Roche Diagnostics).
Statistical analysis. GraphPad Prism software package
(version 7.00) was used to investigate the statistical signicance
of bead- based immunoassay data, while SomaSuite (version
1.0) was used to analyze aptamer- based assay data. The ability
of quantied proteins/peptides to predict the diagnosis (PsA or
RA) for individual patients was assessed using the random forest
package in R (version 3.3.2). The most important variables in pro-
viding the receiver operating characteristic (ROC) area under the
curve (AUC) were selected using the variable importance index,
and the Gini decrease in impurity was used to assess the impor-
tance of each variable. All AUC values were obtained using the
ROC R package.

BIOMARKERS DIFFERENTIATE PsA FROM RA
|
85
RESULTS
Patient sample characterization and study design.
For the discovery of novel candidate protein biomarkers, serum
samples were collected at baseline from patients with early-
onset, treatment- naive PsA (n = 32) and those with early- onset,
treatment- naive RA (n = 32). Samples from a second independent
cohort (PsA, n = 95; RA, n = 72) were used to conrm the perfor-
mance of the putative markers identied during discovery. While
these PsA and RA patients may have been receiving treatment at
the time of baseline serum sampling, there were similar levels of
active disease (as reected by CRP level, erythrocyte sedimen-
tation rate [ESR], and joint counts) in both patient groups. Key
demographic and clinical characteristics for all patients are sum-
marized in Table 1.
Unbiased nano- LC- MS/MS– based protein analysis.
To investigate differential serum protein expression between
patients with PsA and those with RA, individual serum sam-
ples that had been depleted of high- abundance serum proteins
were analyzed by nano- LC- MS/MS using a Q Exactive Hybrid
Quadrupole- Orbitrap mass spectrometer. A total of 451 proteins
were identied, of which 121 were identied in all 64 individual
serum samples. Univariate analysis was applied to the 121 com-
monly identied proteins, and multivariate analysis was applied to
the complete data set. Univariate analysis (Student’s t- test using
a Benjamini- Hochberg false discovery rate of 0.01) showed that
66 proteins were signicantly differentially expressed between
PsA and RA (Supplementary Table 1, http://onlinelibrary.wiley.
com/doi/10.1002/art.41899/abstract). Unsupervised hierarchical
cluster and principal components analysis performed using these
66 proteins revealed the overall differences/similarities between
serum protein levels in the individual PsA and RA patients; clear
within- group clustering and between group separations were
observed (Figure 2). Random forest analysis of data from 451
proteins identied in the 64 patient samples demonstrated that
patients with PsA and those with RA could be differentiated with
an AUC of 0.94 (Table 2) (ROC plot in Supplementary Figure 1A,
http://onlinelibrary.wiley.com/doi/10.1002/art.41899/abstract).
Table 1. Baseline demographic and clinical characteristics of patients in the discovery, verication, and evaluation cohorts*
Discovery and biomarker verification cohort Independent cohort for biomarker evaluation
Total (n = 64) PsA (n = 32) RA (n = 32) Total (n = 167) PsA (n = 95) RA (n = 72)
Age 43.6 ± 13.3 39.6 ± 11.14† 47.7 ± 14.1 53 ± 8.1 52 ± 6.6 55 ± 9.6
Female, no. (%) 37 (58) 15 (47 ) 22 (69) 89 (53) 51 (54) 38 (53)
Anti- CCP positive, no. (%) 33 (52) 026 ( 81) 49 (29) 1 (1) 48 (67)
RF positive, no. (%) 25 (39) 025 (78) 50 (30) 3 (3) 47 (65)
ESR, mm/hour 19.4 ± 16.8 12.0 ± 8.1‡ 26.7 ± 20.0 NA NA NA
CRP, mg/liter (normal <5) 14.4 ± 19.8 6.6 ± 8.3‡ 22.2 ± 24.6 24.9 ± 30.6 28.2 ± 27.8§ 20 ± 34.0
DAS28- CRP, median (IQR) 4.2 (1.66– 6.88) 3.7 (2.1 – 5. 8 ) 4.9 (1.7– 6.9) NA¶ NA¶ 4. 2 (1.1– 7.6)
TJC, median (IQR) (range 0– 28) 6 (0 – 23) 4 (0 – 20)# 8.5 (0– 23) NA¶ 10 .4 (0 – 3 8) ¶ 8.2 (0– 28)
SJC, median (IQR) (range 0– 28) 2 (0– 12) 1 (0 – 5)‡ 3.5 (0– 12) NA¶ 7.2 (0– 25)¶ 5.2 (0 – 24)
Dact ylitis, no. (%) NA 10 ( 31) NA NA¶ 4 4 (46.3) NA
BMI, kg/m228.1 ± 6.3 27.97 ± 6.3 28.24 ± 6.3 28.0 ± 8.6 30.0 ± 10.6‡ 27.2 ± 5.1
PASI, median (range) NA 3 . 35 ( 0 – 27.7 ) NA NA 2. 2 (0–14) NA
* Except where indicated otherwise, values are the mean ± SD. anti- CCP = anti– cyclic citrullinated peptide; RF = rheumatoid factor; ESR =
erythrocyte sedimentation rate; NA = not available; IQR = interquartile range; BMI = body mass index; PASI = Psoriasis Area and Severity
Index.
† P < 0.05 versus rheumatoid arthritis (RA) patients.
‡ P < 0.01 versus RA patients.
§ P < 0.0001 versus RA patients.
¶ For the validation cohort, 68 and 66 joints were counted for the tender joint count (TJC) and swollen joint count (SJC), respectively, in the
psoriatic arthritis (PsA) group, and therefore the Disease Activity Score in 28 joints using the C- reactive protein level (DAS28- CRP) could not
be calculated.
# P < 0.001 versus RA patients.
Figure 2. Association of protein signatures with diagnosis of
psoriatic arthritis (PsA) or rheumatoid arthritis (RA). A, Unsupervised
hierarchical cluster analysis. B, Supervised hierarchical cluster
analysis. C, Principal components analysis. Plots were generated
for differentially expressed proteins between PsA patients (n = 30)
and RA patients (n = 30). P ≤ 0.01 by Benjamin- Hochberg false
discovery rate.

Mc ARDLE ET AL
86
|
Taken together, these data strongly suggest that there is a differ-
ence in the serum protein proles between newly diagnosed PsA
patients and RA patients. The top 50 proteins providing the AUC
are listed in Supplementary Table 2 (http://onlinelibrary.wiley.com/
doi/10.1002/art.41899/abstract).
Aptamer- and bead- based targeted protein anal-
ysis. To extend the breadth and depth of proteome coverage
afforded by nano- LC- MS/MS, serum samples were subjected
to analysis using 2 complementary protein measurement plat-
forms. Aptamer- based analysis supported the quantication of
1,129 proteins in a subset of the patient samples for PsA (n = 18)
and RA (n = 18). Univariate analysis revealed that 175 proteins
were signicantly differentially expressed between PsA and RA
patients (Supplementary Table 3, http://onlinelibrary.wiley.com/
doi/10.1002/art.41899/abstract). Multivariate analysis of the data
obtained from the aptamer- based analysis revealed that it was
possible to discriminate PsA from RA with an AUC of 0.73 (Table 2)
(ROC plot in Supplementary Figure 1B, http://onlinelibrary.wiley.
com/doi/10.1002/art.41899/abstract).
Based largely on their known importance in PsA and RA (3),
48 proteins were selected for analysis using in-house– developed
multiplexed bead- based immunoassays (10). Of the 48 proteins
targeted, 23 were identied in every sample. T- tests revealed that
4 proteins (IL- 18 [P ≤ 0.001], IL- 18 binding protein [P ≤ 0.05],
hepatocyte growth factor [P ≤ 0.05], and tumor necrosis fac-
tor receptor superfamily member 6 [P ≤ 0.05]) were differentially
expressed between PsA and RA samples (Supplementary Figure 2,
http://onlinelibrary.wiley.com/doi/10.1002/art.41899/abstract).
Random forest analysis of the bead- based immunoassay data
showed that patients could be segregated with an AUC of 0.69
(Table 2 and Supplementary Figure 1C). Compared to the nano-
LC- MS/MS analysis, the candidate protein biomarker discovery
by both aptamer- based and bead- based assays yielded data sets
with reduced predictive power, and therefore the subsequent eval-
uation process was streamlined to focus only on proteins identi-
ed by nano- LC- MS/MS.
LC-MRM verication of nano-LC-MS/MS–identied
biomarkers. MRM is a targeted MS technology that is increas-
ingly used to support candidate biomarker evaluation following
LC- MS/MS and other protein discovery approaches. Both the cost
of MRM analysis and the time required to develop and optimize
MRM assays are considerably less than antibody- based methods
(27). For these and other reasons, MRM- based measurement of
the nano- LC- MS/MS– identied proteins represents an attractive
approach for verication and evaluation of their biomarker per-
formance. The multiplexing capabilities afforded by MRM facili-
tated the development of an assay that included the top- ranking
discriminatory candidate proteins from univariate and multivariate
analysis of the nano- LC- MS/MS discovery data described above,
but also allowed for the inclusion of additional proteins identied
previously during studies of pooled patient samples (data not
shown). A total of 233 proteins represented by 735 peptides and
3,735 transitions (5 per peptide) were brought forward for MRM
assay development. Of the 233 proteins brought forward, it was
possible to develop assays for 150 of them, represented by 299
peptides. The remaining candidates could not be detected repro-
ducibly in crude serum. Of the 50 proteins listed in Supplementary
Table 2, 33 were included in the assay.
This MRM assay panel was then used to measure the can-
didate proteins in 60 patient samples from the discovery cohort.
It is noteworthy that to minimize any technical bias, both the pre-
analytical processing and MRM analysis were undertaken in a
randomized manner. Random forest analysis revealed that using
this MRM assay panel it was possible to distinguish PsA from RA
with an AUC of 0.79 (Figure 3A). While this initial work was in pro-
gress, we independently found an additional 23 candidate bio-
marker proteins to be capable of identifying other forms of IA (28).
MRM assays for these proteins were developed and added to the
initial MRM assay panel, yielding a new total number of proteins
of 173 (represented by 334 peptides). This expanded panel was
used to measure candidate proteins in an independent evaluation
cohort of 95 PsA patients and 72 RA patients (Table 1). Seven syn-
thetic isotopically labeled (SIL) peptides were incorporated into the
assay to control for potential analytical variation. Summed inten-
sity values from the SIL peptides were used to normalize patient
data. Random forest analysis revealed that PsA patients could be
separated from those with RA with an AUC of 0.85 (Figure 3B).
The proteins ranked as most important in providing the AUC val-
ues are reported in Supplementary Table 4 (http://onlinelibrary.
wiley.com/doi/10.1002/art.41899/abstract).
Table 2. Determination of protein signatures to predict diagnosis in
patients with early PsA and those with RA*
Platform No.
Correctly
predicted/total AUC
LC- MS/MS 60 55/60 0.94
Aptamer- based immunoassay 36 26/36 0.73
Bead- based immunoassay 64 43/64 0.69
* Area under the curve (AUC) values were generated using predicted
probabilities from the random forest model used to discriminate
between the groups. PsA = psoriatic arthritis; RA = rheumatoid
arthritis; LC- MS/MS = liquid chromatography mass spectrometry.
Figure 3. Receiver operating characteristic curve for performance
of protein signatures in the discovery cohort (n = 30 psoriatic arthritis
[PsA] patients and 30 rheumatoid arthritis [RA] patients) (A) and in
the independent evaluation cohort (n = 95 PsA patients and 72 RA
patients) (B). AUC = area under the curve.

BIOMARKERS DIFFERENTIATE PsA FROM RA
|
87
The data demonstrate clear overlap between proteins used
to distinguish PsA patients from RA patients included in the dis-
covery and verication cohorts. The differential expression levels
of these overlapping proteins are illustrated in Figure 4. To this
end, α2- HS glycoprotein, α1- antichymotrypsin, haptoglobin,
haptoglobin- related protein, and RF C6 light chain (Vκ1) were
found to be signicantly up- regulated in RA patients compared to
PsA patients when measured by MRM. Alpha- 1- acid glycoprotein
and coagulation factor XI were also found to be up- regulated in
RA compared to PsA during both biomarker verication and the
evaluation phase, but the observation only reached signicance
during the evaluation phase. This highlights the value in developing
MRM assays for large panels of candidate proteins and evaluating
them using additional independent patient cohorts. In the case of
thrombospondin 1 (TSP- 1), the protein was found to be slightly up-
regulated in RA patients during verication in the initial discovery
cohort but was signicantly up- regulated in PsA patients during
the subsequent validation stage. It is evident that the potential PsA
versus RA discriminatory role of this protein will require continued
evaluation using additional independent cohorts.
Taken together, these observations provide support for the
strategy we adopted, i.e., to use discovery experiments to gener-
ate an extensive panel of candidates and to use analytically robust
MRM assays to verify their performance (using the initial discovery
cohort), with a separate cohort of patients for evaluation. It is note-
worthy that all samples used here were from patients who under-
went detailed and expert clinical evaluation. It is also apparent
that the strategy can be used to develop an initial classier which
can be tested and further developed to improve the performance
of the predictive algorithm. This ongoing evolution of the MRM
assay panel and associated machine learning algorithms repre-
sent a new and powerful approach to biomarker development.
Figure 4. Protein expression changes in PsA and RA, as measured by multiple reaction monitoring (MRM). Eight proteins contributing to the
AUC generated during target biomarker verication (AUC 0.79) and evaluation (AUC 0.85) show concordant expression changes in independent
cohorts. A, During the initial verication phase, α1- acid glycoprotein 1 (A1AG), coagulation factor XI (FA11), and thrombospondin 1 (TSP- 1)
were not signicantly differently expressed between PsA and RA patients. Proteins α2- glycoprotein (A2AGL) (P < 0.006), α1- antichymotrypsin
(AACT) (P < 0.020), haptoglobin (HPT) (P < 0.001), and haptoglobin- related protein (HPTR) (P < 0.015) were signicantly up- regulated in RA. B,
During a subsequent evaluation phase, α1- acid glycoprotein 1 (P < 0.00001), α2- glycoprotein (P < 0.00001), α1- antichymotrypsin (P < 0.00001),
haptoglobin (P < 0.0001), haptoglobin- related protein (P < 0.00001), Vκ1 (P < 0.0001), and coagulation factor XI were signicantly up- regulated
in RA, while TSP- 1 was signicantly up- regulated in PsA (P < 0.00001). C, MRM and mass spectrometry spectrum for C- reactive protein (CRP)
levels are shown. D, CRP levels analyzed by enzyme- linked immunosorbent assay (ELISA) (P ≤ 0.009) and MRM (P ≤ 0.006) are shown. E,
Pearson’s correlation between ELISA and MRM measurements of CRP levels (R2 = 0.8345) is shown. See Figure 3 for other denitions.

Mc ARDLE ET AL
88
|
Finally, there are at least 2 potential routes to implementing a multi-
plexed protein biomarker panel in the clinical setting. One is to use
MRM assays and the other to develop antibody- based assays
for the proteins of interest. To explore the extent to which MRM
data may align with ELISA, we compared our MRM data on CRP
levels with results obtained by standard clinical laboratory ELISA.
MRM measurements were compared to the ELISA measurements
in the 60 samples from the discovery set. It was not surprising
to nd that serum CRP levels were signicantly up- regulated in
patients with RA compared to those with PsA when measured by
both ELISA (P ≤ 0.005) and MRM (P ≤ 0.001) (Figure 4D). Interest-
ingly, the CRP values from both platforms were strongly correlated
(R2 = 0.8345) (Figure 4E), indicating that protein (peptide) mea-
surements obtained by MRM can provide values similar to those
obtained by existing immunoassays.
DISCUSSION
PsA is a complex disease with diverse manifestations; the
clinical features observed in individuals with PsA often vary sub-
stantially but can overlap with other diseases. Differentiating
between PsA and RA can be clinically challenging because of the
similarities in their clinical presentation (29). It is increasingly evi-
dent that making an accurate diagnosis is important in order to
determine which therapeutic strategy to adopt to optimize clinical
and radiographic outcomes (30). With no diagnostic laboratory
test available, the diagnosis is clinical: it depends on the skills and
knowledge of the assessor and is commonly based on the pres-
ence of inammatory musculoskeletal disease in a patient with
skin/nail psoriasis and in the absence of RF (31). However, the
lack of clear denitions for dermatologists and general practition-
ers for inammatory musculoskeletal disease, coupled with inad-
equate training in musculoskeletal examination techniques, leads
to diagnostic uncertainty and delay. As many as 30% of psoriasis
patients visiting dermatology practices may have undiagnosed
PsA (32). A diagnostic delay of >6 months is not uncommon, and
this contributes to poor radiographic and functional outcomes
(33,34).
There is a critical need to differentiate PsA from other forms of
IA, including RA, and to develop and disseminate new approaches
for the objective and sensitive diagnosis of PsA. This is especially
important at the early stages of less differentiated disease, when
a clear diagnosis and the establishment of disease- appropriate
therapy may have the most impact in improving outcomes. Only a
few studies have investigated whether there are biomarkers which
discriminate between PsA and RA. In one study involving syno-
vial tissue, messenger RNA for vascular endothelial growth factor
and angiopoietin 2 were elevated in PsA patients compared to RA
patients (35). However, obtaining a synovial biopsy specimen is an
invasive procedure, and the discomfort, time, and cost associated
with tissue sampling makes it highly undesirable for use in routine
clinical practice (35,36). More recently, Siebert et al identied 170
urinary peptides that discriminated between patients with long-
standing PsA and those with other arthropathies, including early
RA, with an AUC of 0.97 (37). These ndings are very promising,
but urine collection is especially vulnerable to physiologic variation
arising from diet and liquid intake. Additionally, urine tends to be
a very diluted matrix high in salt and low in protein concentration.
Thus, in the absence of stepwise workows for sample concen-
tration and clean- up, the quantication of proteins in urine can
prove difcult as a result of interfering signals present in the matrix
(38).
Serum is well recognized as a suitable sample for biomarker
discovery, not least because proteins are shed from relevant
affected tissues into the circulation, but also because it is readily
obtained under standardized operating procedures (39). Thus, we
used serum samples analyzed by 3 proteomic platforms (nano-
LC- MS/MS, aptamer- based assays, and bead- based assays).
Each platform is capable of measuring a limited but comple-
mentary range of proteins present at different abundance levels.
This approach was adopted in order to maximize coverage of
the serum proteome, and to date it is the most comprehensive
analysis of the serum proteome in patients with PsA and those
with RA. Although 3 platforms were used to identify putative bio-
markers, the data from the unbiased nano- LC- MS/MS analysis
proved to be more discriminatory compared to the data from the
bead- based and aptamer- based platforms. A potential reason for
this is that LC- MS/MS analysis allows for unbiased discovery of
biomarkers, whereas the other approaches are limited by having
xed panels of protein markers. Furthermore, the aptamer- based
platform uses a single aptamer to capture proteins, thus poten-
tially reducing the specicity of readouts (40). It is also possible
that the smaller number of patient samples used in the aptamer-
based experiments may have constrained the statistical power
of the analysis. With respect to the bead- based immunoassay,
the 48 carefully selected proteins we measured may not have
included key candidate cytokines and chemokines which could
support the differentiation between PsA and RA. The proteins
were selected based on their known importance in the pathogen-
esis of PsA and RA, but the panel was limited by the availability of
proteins measurable with the in- house assay.
With no compelling evidence to justify the time and cost
required to develop further multiplex antibody- based and/or
aptamer- based assays, we instead focused on the nano- LC- MS/
MS data and performed follow- up studies using MRM. MRM is an
excellent tool for supporting large- scale, multiprotein biomarker
studies. It is typically used to narrow an initial list of candidate pro-
teins derived from discovery experiments to the subset that may
truly address the clinical question under study (41). MRM analysis
is performed using triple- quadrupole mass spectrometers, which
inherently have higher sensitivity and greater linear dynamic range
than the Orbitrap mass spectrometer used in the discovery exper-
iments here. This boost in sensitivity facilitates the detection of
low- abundant proteins in complex samples and therefore reduces

BIOMARKERS DIFFERENTIATE PsA FROM RA
|
89
the need for sample pre- enrichment steps. Thus, MRM sup-
ports more robust workows as well as time- and cost- effective
assay development compared to traditional antibody- based
approaches. MRM is frequently less sensitive than an equivalent
immunoassay, and it was for this reason that we did not initially
attempt to develop MRM assays for putative markers identied
only by the aptamer- based or the bead- based analysis (17,42).
The development of MRM immunoassays for these candidate
biomarker proteins represents an obvious way in which improving
the performance of the existing panel could be explored (43).
In the 2 phases of MRM analysis described here, it was
especially interesting to note that a subpanel of 8 proteins
(leucine- rich α2- glycoprotein, α1- antichymotrypsin, haptoglobin,
haptoglobin- related protein, RF C6 light chain, α1- acid glyco-
protein 1, coagulation factor XI, and TSP- 1) that were identied
as highly discriminatory during the initial verication phase were
again conrmed as highly discriminatory during the second eval-
uation phase. Follow- up t- test analysis was performed on this set
of proteins, and 7 of 8 proteins were found to be up- regulated in
RA compared to PsA during both phases of analysis. TSP- 1 was
found to be signicantly up- regulated in PsA compared RA during
the second phase, whereas no signicant difference was observed
in initial verication. This discordance may relate to differences in
the number of patients included in the 2 phases, or it may relate to
the differences in the patients included; patients in the initial phase
had early- onset, treatment- naive disease, while those included in
the second phase had longer- standing disease and were receiving
therapy. This highlights, in part, the advantage of maintaining large
panels of proteins for ongoing evaluation in patient cohorts.
Further analysis of this 8- protein subpanel was carried out
using a web- based resource “Search Tool for the Retrieval of
Interacting Genes/Proteins” (https://strin g- db.org/cgi/netwo rk.pl),
revealing the biologic functions of these 8 markers of interest (Sup-
plementary Table 5, http://onlinelibrary.wiley.com/doi/10.1002/art.
41899/abstract). It is interesting to note that this panel is enriched
for proteins functionally involved in structural remodeling, angi-
ogenesis, homeostasis, and transportation. This perhaps is not
surprising since PsA and RA are characterized by an increase in
bone turnover and dysregulated angiogenesis. The radiographic
features in PsA and RA can be quite different, with bony erosion
observed in both conditions but osteoproliferation only seen in
PsA (3). In the context of this investigation, it was not unanticipated
that markers of structural remodeling contributed to an algorithm
discriminating between individuals with PsA and those with RA.
Here, we demonstrated that a major advantage of using MRM is
that it allows the investigator to rapidly adapt a panel to include
new candidate biomarkers. Our CRP assay that was developed
using MRM over a few days also showed values highly correlated
with those generated by ELISA.
Our study has several strengths, including the comprehen-
sive and logical approach to biomarker development. Limitations
include the modest number of patient samples in both study
phases as well as the absence of healthy and disease controls.
Differentiating between PsA and RA is the focus of the current
study, but it is not the only challenge faced by clinicians, as it can
also be challenging to distinguish PsA from other arthropathies
and from patients who have skin psoriasis only (14). This certainly
represents a future objective, and assessing this biomarker panel
in the appropriate additional cohorts is a critical next step. It is
noteworthy that the independent cohort included in the second
phase of evaluation included patients that had longstanding dis-
ease compared to the discovery cohort, which included those
with early- onset disease. Despite this, the 2 cohorts shared similar
levels of active disease, as reected by CRP level, ESR, and joint
counts (Table 1). However, it should be noted that the Disease
Activity Score using the CRP level (DAS28- CRP) (44) was used
as a disease activity measure in the PsA discovery cohort. This
is not recommended, since it does not reect the 68- joint counts
recommended for the disease. Notwithstanding this, the DAS28-
CRP results show that while lower in PsA, the mean values are not
signicantly different between the 2 diseases.
It is fair to say that the patients included in this study are rep-
resentative of those attending IA clinics. We believe that obtaining
data and samples from real- world conditions is critically important
if our assay is to consistently segregate PsA from RA regardless
of disease duration, disease activity, treatment, or comorbidities.
The performance of the biomarker panel may reect a genuine
difference in the protein prole between PsA and RA patients, but
further work in a larger number of patient samples is needed. It will
also be necessary to examine the performance of the panel in dis-
tinguishing PsA from other forms of IA and from healthy individuals.
It should be noted that all PsA patients included in both
the discovery and verication cohorts met the CASPAR criteria,
which was required for inclusion. Therefore, it was not possible
in this study to compare the performance of the biomarker panel
to that of CASPAR criteria or to test whether a combination of
CASPAR criteria and biomarkers is more useful. We intend to
address this in a prospective study of psoriasis patients who
are being followed up for the development of PsA or in a cohort
of patients with early undifferentiated IA. Finally, although non-
inammatory disease controls were not included in our present
analysis, it is worth highlighting research by Chandran et al that
identied differences in serum proteins in patients with PsA com-
pared to patients with osteoarthritis (45) and patients with psoria-
sis (46). The protein markers identied in these studies are prime
candidates that should be included in future generations of MRM
panel assays. At present, there is no diagnostic test for PsA and
as a result, the diagnosis is often late or missed, resulting in func-
tional consequences for the patient (12,47). With at least 20% of
the patients referred to early arthritis clinics diagnosed as having
PsA, there is an urgent need to develop a test to support early
detection of this disease (31).
In conclusion, the work described here represents a signicant
contribution toward the development of such a test. Fundamental

Mc ARDLE ET AL
90
|
next steps have been outlined, and the MRM approach is ideally
suited to support the large- scale studies required to develop and
validate a robust panel of distinguishing biomarkers. We believe that
with further development it will be possible to establish a diagnos-
tic test for PsA that will reduce diagnostic delay, inform treatment
selection, and improve both short- term and long- term outcomes.
ACKNOWLEDGMENT
The authors would like to thank Prof. Anthony G. Wilson (Arthritis
Ireland Chair of Rheumatology, UCD Centre for Arthritis Research, School
of Medicine, Conway Institute, Beleld, Dublin 4). Prof. Wilson made a
fundamental contribution to our study by providing the RA cohort used in
our validation analysis.
AUTHOR CONTRIBUTIONS
All authors were involved in drafting the article or revising it critically
for important intellectual content, and all authors approved the nal version
to be published. Dr. Mc Ardle had full access to all of the data in the study
and takes responsibility for the integrity of the data and the accuracy of
the data analysis.
Study conception and design. Mc Ardle, Szentpetery, Hernandez,
Parnell, FitzGerald, Pennington.
Acquisition of data. Mc Ardle, Kwasnik, Szentpetery, de Jager, de Roock,
FitzGerald, Pennington.
Analysis and interpretation of data. Mc Ardle, Kwasnik, Szentpetery,
Hernandez, Parnell, de Jager, de Roock, FitzGerald, Pennington.
REFERENCES
1. Ogdie A, Langan S, Love T, Haynes K, Shin D, Seminara N, et al.
Prevalence and treatment patterns of psoriatic arthritis in the UK.
Rheumatology (Oxford) 2013;52:568– 75.
2. Ogdie A, Weiss P. The epidemiology of psoriatic arthritis [review].
Rheum Dis Clin North Am 2015;41:545– 68.
3. Mc Ardle A, Flatley B, Pennington SR, FitzGerald O. Early biomark-
ers of joint damage in rheumatoid and psoriatic arthritis [review].
Arthritis Res Ther 2015;17:141.
4. McArdle A, Pennington SR, FitzGerald O. Clinical features of pso-
riatic arthritis: a comprehensive review of unmet clinical needs
[review]. Clin Rev Allergy Immunol 2018;55:271– 94.
5. Kane D, Stafford L, Bresnihan B, FitzGerald O. A prospective, clinical
and radiological study of early psoriatic arthritis: an early synovitis
clinic experience. Rheumatology (Oxford) 2003;42:1460– 8.
6. Husted JA, Gladman DD, Farewell VT, Cook RJ. Health- related
quality of life of patients with psoriatic arthritis: a compari-
son with patients with rheumatoid arthritis. Arthritis Rheum
2001;45:151– 8.
7. Sokoll KB, Helliwell PS. Comparison of disability and quality of life in
rheumatoid and psoriatic arthritis. J Rheumatol 2001;28:1842– 6.
8. Lindqvist UR, Alenius GM, Husmark T, Theander E, Holmström G,
Larsson PT, on behalf of the Psoriatic Arthritis Group of the Society
for Rheumatology. The Swedish Early Psoriatic Arthritis Register–
2- year followup: a comparison with early rheumatoid arthritis.
JRheumatol 2008;35:668– 73.
9. Lee S, Mendelsohn A, Sarnes E. The burden of psoriatic arthritis:
a literature review from a global health systems perspective. P T
2010;35:680– 9.
10. McArdle A, Butt AQ, Szentpetery A, de Jager W, de Roock S,
FitzGerald O, et al. Developing clinically relevant biomarkers in
inammatory arthritis: a multiplatform approach for serum candidate
protein discovery. Proteomics Clin Appl 2016;10:691– 8.
11. Ritchlin CT, Kavanaugh A, Gladman DD, Mease PJ, Helliwell P,
Boehncke WH, et al. Treatment recommendations for psoriatic
arthritis [review]. Ann Rheum Dis 2009;68:1387– 94.
12. Kirkham B, de Vlam K, Li W, Boggs R, Mallbris L, Nab HW, et al.
Early treatment of psoriatic arthritis is associated with improved
patient- reported outcomes: ndings from the etanercept PRESTA
trial. Clin Exp Rheumatol 2015;33:11– 9.
13. Yago T, Nanke Y, Kawamoto M, Kobashigawa T, Yamanaka H,
Kotake S. IL- 23 and Th17 disease in inammatory arthritis [review].
J Clin Med 2017;6:81.
14. Butt AQ, McArdle A, Gibson DS, FitzGerald O, Pennington SR.
Psoriatic arthritis under a proteomic spotlight: application of novel
technologies to advance diagnosis and management [review]. Curr
Rheumatol Rep 2015;17:35.
15. Taylor W, Gladman D, Helliwell P, Marchesoni A, Mease P, Mielants
H, on behalf of the CASPAR Study Group. Classication criteria for
psoriatic arthritis: development of new criteria from a large interna-
tional study. Arthritis Rheum 2006;54:2665– 73.
16. Coates LC, Conaghan PG, Emery P, Green MJ, Ibrahim G,
MacIver H, et al. Sensitivity and specicity of the classication of
psoriatic arthritis criteria in early psoriatic arthritis. Arthritis Rheum
2012;64:3150– 5.
17. Chandran V. Spondyloarthritis: CASPAR criteria in early psoriatic
arthritis [review]. Nat Rev Rheumatol 2012;8:503– 4.
18. Gibson DS, Rooney ME, Finnegan S, Qiu J, Thompson DC, Labaer
J, et al. Biomarkers in rheumatology, now and in the future [review].
Rheumatology (Oxford) 2012;51:423– 33.
19. Szentpetery A, Herffernan E, Haroon M, Kilbane M, Gallagher P,
McKenna MJ, et al. Striking difference of periarticular bone density
change in early psoriatic arthritis and rheumatoid arthritis following
anti- rheumatic treatment as measured by digital X- ray radiogramme-
try. Rheumatology (Oxford) 2016;55:891– 6.
20. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson D, Bingham CO III,
et al. 2010 rheumatoid arthritis classication criteria: an American
College of Rheumatology/European League Against Rheumatism
collaborative initiative. Arthritis Rheum 2010;62: 2569– 81.
21. Rappsilber J, Ishihama Y, Mann M. Stop and go extraction tips
for matrix- assisted laser desorption/ionization, nanoelectros-
pray, and LC/MS sample pretreatment in proteomics. Anal Chem
2003;75:663– 70.
22. Turriziani B, Garcia- Munoz A, Pilkington R, Raso C, Kolch W,
von Kriegsheim A. On- beads digestion in conjunction with data-
dependent mass spectrometry: a shortcut to quantitative and
dynamic interaction proteomics. Biology (Basel) 2014;3:320– 32.
23. Cox J, Neuhauser N, Michalski A, Scheltema RA, Olsen JV, Mann M.
Andromeda: a peptide search engine integrated into the MaxQuant
environment. J Proteome Res 2011;10:1794– 805.
24. MacLean B, Tomazela DM, Shulman N, Chambers M, Finney GL,
Frewen B, et al. Skyline: an open source document editor for creat-
ing and analyzing targeted proteomics experiments. Bioinformatics
2010;26:966– 8.
25. Molecular and Cellular Proteomics. Targeted Mass Spec Guidelines.
URL: https://www.mcpon line.org/guide lines - publi catio n- manus cript
s- descr ibing - devel opmen t- and- appli catio n- targe ted- mass.
26. Grant RP, Hoofnagle AN. From lost in translation to paradise found:
enabling protein biomarker method transfer by mass spectrometry.
Clin Chem 2014;60:941– 4.
27. Picotti P, Bodenmiller B, Aebersold R. Proteomics meets the scien-
tic method. Nat Methods 2013;10:24– 7.
28. Gohar F, McArdle A, Jones M, Callan N, Hernandez B, Kessel C,
et al. Molecular signature characterisation of different inammatory
phenotypes of systemic juvenile idiopathic arthritis. Ann Rheum Dis
2019;78:1107– 13.

BIOMARKERS DIFFERENTIATE PsA FROM RA
|
91
29. Haroon M, FitzGerald O. Psoriatic arthritis: complexities, comorbidi-
ties and implications for the clinic [review]. Expert Rev Clin Immunol
2016;12:405– 16.
30. Merola JF, Espinoza LR, Fleischmann R. Distinguishing rheumatoid
arthritis from psoriatic arthritis. RMD Open 2018;4:e000656.
31. Coates LC, Helliwell PS. Psoriatic arthritis: state of the art review
[review]. Clin Med (Lond) 2017;17:65– 70.
32. Haroon M, Kirby B, FitzGerald O. High prevalence of psoriatic
arthritis in patients with severe psoriasis with suboptimal performance of
screening questionnaires. Ann Rheum Dis 2013;72: 736– 40.
33. Haroon M, Gallagher P, FitzGerald O. Diagnostic delay of more than
6 months contributes to poor radiographic and functional outcome
in psoriatic arthritis. Ann Rheum Dis 2015;74:1045– 50.
34. Tillett W, Jadon D, Shaddick G, Cavill C, Korendowych E, de Vries
CS, et al. Smoking and delay to diagnosis are associated with
poorer functional outcome in psoriatic arthritis. Ann Rheum Dis
2013;72:1358– 61.
35. Fearon U, Griosios K, Fraser A, Reece R, Emery P, Jones PF, et al.
Angiopoietins, growth factors, and vascular morphology in early
arthritis. J Rheumatol 2003;30:260– 8.
36. Ilie M, Hofman P. Pros: can tissue biopsy be replaced by liquid
biopsy? [editorial]. Transl Lung Cancer Res 2016;5:420– 3.
37. Siebert S, Porter D, Paterson C, Hampson R, Gaya D, Latosinska A,
et al. Urinary proteomics can dene distinct diagnostic inammatory
arthritis subgroups. Sci Rep 2017;7:40473.
38. Chiu M, Lawi W, Snyder SI, Wong PK, Liao JC, Gau G. Matrix
effects— a challenge toward automation of molecular analysis. J Lab
Autom 2010;15:233– 42.
39. Wilson R. Sensitivity and specicity: twin goals of proteomics
assays. Can they be combined? [review]. Expert Rev Proteomics
2013;10:135– 49.
40. Joshi A, Mayr M. In aptamers they trust: the caveats of the
SOMAscan biomarker discovery platfrom from SomaLogic.
Circulation 2018;138:2482– 5.
41. Carr SA, Abbatiello SE, Ackermann BL, Borchers C, Domon B,
Deutsch EW, et al. Targeted peptide measurements in biology
and medicine: best practices for mass spectrometry- based assay
development using a t- for- purpose approach. Mol Cell Proteomics
2014;13:907– 17.
42. Lange V, Picotti P, Domon B, Aebersold R. Selected reaction monitor-
ing for quantitative proteomics: a tutorial. Mol Syst Biol 2008;4:222.
43. Yassine H, Borges CR, Schaab MR, Billheimer D, Stump C, Reaven
P, et al. Mass spectrometric immunoassay and MRM as targeted
MS- based quantitative approaches in biomarker development:
potential applications to cardiovascular disease and diabetes
[review]. Proteomics Clin Appl 2013;7:528– 40.
44. Prevoo ML, van ’t Hof MA, Kuper HH, van Leeuwen MA, van de
Putte LB, van Riel PL. Modied disease activity scores that include
twenty- eight– joint counts: development and validation in a prospec-
tive longitudinal study of patients with rheumatoid arthritis. Arthritis
Rheum 1995;38:44– 8.
45. Chandran V, Abji F, Perruccio AV, Gandhi R, Li S, Cook RJ, et al.
Serum- based soluble markers differentiate psoriatic arthritis from
osteoarthritis. Ann Rheum Dis 2019;78:796– 801.
46. Chandran V, Cook RJ, Edwin J, Shen H, Pellett FJ, Shanmugarajah
S, et al. Soluble biomarkers differentiate patients with psoriatic arthri-
tis from those with psoriasis without arthritis. Rheumatology (Oxford)
2010;49:1399– 405.
47. Sorensen J, Hetland ML, on behalf of all departments of rheu-
matology in Denmark. Diagnostic delay in patients with rheuma-
toid arthritis, psoriatic arthritis and ankylosing spondylitis: results
from the Danish nationwide DANBIO registry. Ann Rheum Dis
2015;74:e12.