ARTHRITIS & RHEUMATISM

Vol. 50, No. 10, October 2004, pp 3093–3103

DOI 10.1002/art.20555
© 2004, American College of Rheumatology

Interaction Between RANKL and HLA–DRB1 Genotypes

May Contribute to Younger Age at Onset of
Seropositive Rheumatoid Arthritis in an Inception Cohort

Hui Wu,1 Dinesh Khanna,2 Grace Park,2 Vivian Gersuk,3 Gerald T. Nepom,3
Weng Kee Wong,2 Harold E. Paulus,2 and Betty P. Tsao,2
for the Western Consortium of Practicing Rheumatologists

Objective. To determine whether the RANKL and
HLA–DRB1 “shared epitope” (SE) genotypes contribute
to the development of rheumatoid arthritis (RA).
Methods. We studied 237 patients with early RA
(within 15 months of symptom onset) who were seropos-
itive for rheumatoid factor. HLA–DRB1 genotyping was
performed using the polymerase chain reaction (PCR)–
based oligonucleotide probe assay. RANKL polymor-
phisms were analyzed using PCR pyrosequencing for
Supported in part by grants from the NIH (P60-AR-36834),
the Southern California Chapter of the Arthritis Foundation, and the
Paxson Family Foundation. Dr. Khanna’s work was supported by a
Centocor Health Outcomes in Rheumatic Diseases Fellowship.
1 (53 ⴞ 12.5 years; P < 0.0001) and was 17 years younger
Hui Wu, MD: University of California, Los Angeles, and
Ren Ji Hospital, Shanghai Second Medical University, Shanghai,
China; 2Dinesh Khanna, MD, MS (current address: University of
Cincinnati, Cincinnati, Ohio), Grace Park, MPH, Weng Kee Wong,
PhD, Harold E. Paulus, MD, Betty P. Tsao, PhD: University of
California, Los Angeles; 3Vivian Gersuk, PhD, Gerald T. Nepom, MD,
PhD: Benaroya Research Institute at Virginia Mason, Seattle, Wash-
ington.
Members of the Western Consortium of Practicing Rheuma-
tologists are as follows: Drs. Robert Shapiro, Maria W. Greenwald, H.
Walter Emori, Fredrica E. Smith, Craig W. Wiesenhutter, Charles
Boniske, Max Lundberg, Anne MacGuire, Jeffry Carlin, Robert
Ettlinger, Michael H. Weisman, Elizabeth Tindall, Karen Kolba,
George Krick, Melvin Britton, Rudy Greene, Ghislaine Bernard
Medina, Raymond T. Mirise, Daniel E. Furst, Kenneth B. Wiesner,
Robert F. Willkens, Kenneth Wilske, Karen Basin, Robert Gerber,
Gerald Schoepflin, Marcia J. Sparling, George Young, Philip J. Mease,
Ina Oppliger, Douglas Roberts, J. Javier Orozco Alcala, John Seaman,
Martin Berry, Ken J. Bulpitt, Grant Cannon, Gregory Gardner, Allen
Sawitzke, Andrew Lun Wong, Daniel O. Clegg, Timothy Spiegel,
Wayne Jack Wallis, Mark Wener, Robert Fox.
Drs. Wu and Khanna contributed equally to this work.
Address correspondence and reprint requests to Betty P.
Tsao, PhD, Division of Rheumatology, Department of Medicine,
Rehabilitation Center, Room 32-59, 1000 Veteran Avenue, UCLA
School of Medicine, Los Angeles, CA 90095-1670. E-mail:
Btsao@mednet.ucla.edu.
Submitted for publication February 26, 2004; accepted in
revised form June 22, 2004.
3093
SNP1 and fluorescence-based PCR for the presence or
absence of the TAAA insertion.
Results. The presence of SE-containing DRB1*04
alleles was associated with an earlier age at RA onset
(mean ⴞ SD 47 ⴞ 12.7 years versus 53 ⴞ 12.5 years in
SE– patients; P ⴝ 0.0004). The 2 novel RANKL polymor-
phisms were in strong linkage disequilibrium (P <
0.0001) and were associated with earlier ages at disease
onset (e.g., for the CC versus CT/TT genotypes, 44 ⴞ
13.5 years versus 51 ⴞ 12.7 years; P ⴝ 0.0080). The
mean age at disease onset in SEⴙ patients with the
RANKL-CC genotype (35 ⴞ 7.2 years) was a mean of 18
years younger than in SE– patients with RANKL-CT/TT
than in SE– patients with RANKL-CC (52 ⴞ 13.2 years;
P ⴝ 0.0005). The proportion of patients with both the
SE and RANKL risk alleles was highest (23%) in those
who developed RA during their third decade of life (ages
20–30 years), with a declining trend among those who
developed RA during their fourth (16%), fifth (5%), and
sixth or later (0%) decades. Interestingly, 92% of the
patients diagnosed as having RA between ages 20 and 30
years carried at least 1 of the RA-associated DRB1*04
alleles, suggesting a strong influence of the SE in the
early onset of RA. The majority of patients who devel-
oped RA symptoms in their third to fifth decades (74 of
119 [62%]) carried at least 1 copy of the DRB1*04
alleles; in contrast, fewer than half of the patients who
developed RA in their sixth decade or later (50 of 118
[42%]) had DRB1*04 alleles. RANKL genotypes were
not associated with erosive disease at baseline or with
the yearly progression rate of radiographic joint dam-
age.
Conclusion. This study provides the first evidence
that novel RANKL polymorphisms were associated with
3094
an earlier age at RA onset in SEⴙ, but not SE–, patients
and that an interaction between SE-containing HLA–
DRB1 and RANKL polymorphisms increased the disease
penetrance, resulting in a mean age at RA onset that
was 18–20 years younger. Our results also suggested
genetic differences between patients with early-onset
and those with late-onset RA.
Rheumatoid arthritis (RA) is a systemic auto-
immune disease characterized by chronic inflammation
of the joints, which may lead to structural damage of the
cartilage and bone. Genetic influences in RA have been
supported by cumulative studies (for review, see refs. 1
and 2). The best known and the most dominant genetic
risk factor for RA is the HLA class II region containing
the DRB1 locus. This has been a consistent finding of 4
independent genome scans using family collections from
Europe, the US (2 studies), and the UK (3–6). The
RA-associated DRB1 alleles (DRB1*0101, *0102, *0401,
*0404, *0405, *0408, *0409, *0410, *0413, *0416, *0421,
*1001, *1402, and *1406) share a conserved linear
sequence of amino acids between positions 70 and 74 in
the DR␤1 chain (QKRAA, QRRAA, or RRRAA) of
the HLA–DR␣/␤ heterodimer, which has led to the
“shared epitope” (SE) hypothesis (7).
The risk of RA conferred by different SE-
containing alleles may vary. In addition, there are ethnic
variations in the allele frequency of SE-containing al-
leles associated with RA. In Caucasians, DRB1*04 al-
leles (especially *0401) are the major SE-containing
alleles associated with a higher risk of RA (1,8). The
presence and the dose effect of these DRB1 alleles have
also been associated with early disease onset (8–10),
radiologic erosions (11–18), and extraarticular manifes-
tations (19,20). Most estimates of the HLA component
of the overall genetic risk of RA are ⬍50% (21), which
makes it an important challenge to identify non-HLA
genes that may contribute to the pathogenesis of RA.
RANKL (also known as osteoprotegerin ligand,
TRANCE, and osteoclast differentiation factor), a
member of the tumor necrosis factor (TNF) ligand
family, has been shown to play a pivotal role in inflam-
matory joint diseases (for review, see ref. 22). Cell
membrane–anchored or secreted RANKL binds to
RANK that is expressed on osteoclasts, dendritic cells,
and B and T lymphocytes, resulting in osteoclastogen-
esis, osteoclast activation, dendritic cell survival, lym-
phocyte development, and lymph node organogenesis
(23,24). A natural, soluble decoy receptor, osteoprote-
gerin (OPG), binds both the secreted and cell-bound
forms of RANKL, preventing the interaction and stim-
WU ET AL
ulation of RANK (23). Levels of both OPG and RANKL
are elevated in the sera of RA patients, but are normal-
ized without influencing the ratio of OPG to RANKL
after treatment with anti-TNF (25). The interaction
between RANK and RANKL appears to mediate bone
and cartilage destruction in animal models of arthritis,
and OPG treatment is effective in preventing bone
erosion and has partially beneficial effects on cartilage
destruction (26–28). Taken together, the functional
properties of RANKL make it a logical candidate gene
for RA.
To our knowledge, no studies published thus far
have described an association between genetic polymor-
phisms of RANKL and RA. Single-nucleotide polymor-
phisms (SNPs) that affect the binding of transcription
factors have recently been associated with susceptibility
to multiple autoimmune diseases including RA (29–32).
This provided a rationale for selecting 3 intronic SNPs of
RANKL from the National Center for Biotechnology
Information SNP database that had a predictive alter-
ation in transcription factor affinity between the 2
alleles, using the TFSearch computer program (33).
Among these 3 SNPs (dbSNP nos. rs7334187, rs387856,
and rs922996) the former 2 were not polymorphic in our
samples, leaving 1 SNP (SNP1) that was appropriate for
association study. Since SNP1 is located in intron 7
toward the 3⬘ end of the gene, we selected another
polymorphism (rs5803141; the insertion of TAAA or
none) lying 37.5 kb upstream of the entire 45.3-kb
RANKL for this initial study.
In the present study, we assessed novel RANKL
polymorphisms and HLA–DRB1*04 alleles for their
potential association with RA in an early (⬍15 months
after symptom onset) observational cohort of patients
with seropositive active RA. Since an inception cohort
tends to have a more accurately documented date of RA
onset, we examined whether the age at disease onset
correlated with the DRB1*04 and RANKL genotypes. In
addition, we studied whether RANKL genotypes were
associated with the presence of erosive disease at base-
line and with the rate of radiographic progression of
joint damage.
PATIENTS AND METHODS
Patients. The patients included in this study were a
subset of a group of RA patients who were participating in a
long-term observational study involving the Western Consor-
tium of Practicing Rheumatologists, which is a regional con-
sortium of 29 rheumatology practices in the western United
States and Mexico, as described previously (34–36). The
inclusion criteria for this study were a diagnosis of RA
RANKL AND HLA–DRB1 GENOTYPES AND AGE AT ONSET OF RA
according to the American College of Rheumatology (for-
merly, the American Rheumatism Association) criteria (37)
within 15 months of symptom onset, no previous treatment
with disease-modifying antirheumatic drugs, seropositivity for
rheumatoid factor (a titer ⱖ1:80 or a level ⱖ40 IU/ml), ⱖ6
swollen joints, and ⱖ9 tender joints, and availability of
genomic DNA. Symptom onset was defined as the date when
musculoskeletal symptoms began, provided that these symp-
toms persisted and led to the diagnosis of RA. This study was
approved by appropriate institutional review boards.
Clinical and laboratory assessments. The consortium
rheumatologists assessed the patients’ disease status at study
entry (baseline), 6 months, 1 year, and yearly thereafter. Using
standard methods, the detailed physician’s assessment in-
cluded all of the core set of outcome measures required to
calculate the Disease Activity Score (DAS), including com-
plete tender and swollen joint counts (of 68 tender and 66
swollen joints) and measures of acute-phase reactants, as well
as assessments of global disease activity and pain on 0–100-mm
visual analog scales. The DAS was calculated according to the
published algorithm, using the Ritchie Articular Index, swollen
joint count of 44 joints, and erythrocyte sedimentation rate (in
mm/hour) (38). In addition, assays for rheumatoid factor and
self-report measurements, such as the disability index of the
Health Assessment Questionnaire (HAQ) and the Center for
Epidemiologic Studies Depression Scale (39), were performed
at each visit. At each scheduled physician visit, blood speci-
mens were collected for measurement of C-reactive protein
levels; Westergren erythrocyte sedimentation rates were deter-
mined, when clinically indicated, in the rheumatologist’s office
or local laboratory.
Radiographic assessments. Radiographs of the hands,
wrists, and forefeet were obtained at entry, at 6 months, 1 year,
and yearly thereafter. Standard posteroanterior radiographs
that included both hands and wrists and anteroposterior
radiographs that included both forefeet were obtained in the
rheumatologists’ offices or at the radiology facility they usually
used. All available radiographs were scored by 2 experienced
readers for the presence of erosions (0–5 scale) and joint space
narrowing (0–4 scale), and a total score (sum of the erosion
and joint space narrowing scores) was obtained. Sclerosis or
healing of erosions was not scored.
After a brief “training” session in which the readers
discussed the scoring scale and reviewed a small group of
radiographs to be sure they agreed on the features to be
scored, the radiographs were scored independently using the
method described by Sharp et al (40,41), assessing 17 joints in
each hand and wrist for erosions, 16 joints in each hand and
wrist for joint space narrowing, and 6 joints in each forefoot for
erosions and joint space narrowing. The maximum possible
scores for the hands and wrists were 170 for erosions and 128
for joint space narrowing, with a total score of 298. The
maximum possible scores for the feet were 60 for erosions and
48 for joint space narrowing, with a total score of 108. The
maximum possible scores for all joints assessed (hands, wrists,
and feet bilaterally) were 230 for erosions and 176 for joint
space narrowing, with a total score of 406.
Radiographs were read in patient sets, which had been
randomized and blinded for sequence. The independent scores
of readers 1 and 2 were averaged for each radiograph, and this
average was used for the analyses. Reliability of the readers
3095
was assessed by the intraclass correlation coefficient and the
smallest detectable difference. The intraclass correlation coef-
ficient and the smallest detectable difference for the average of
the 2 readers’ scores was 0.97 and 3.07 for the erosion score,
0.93 and 7.52 for the joint space narrowing score, and 0.90 and
12.71 for the total Sharp score (36).
Because the time interval between pairs of radiographs
varied among patients, radiographic progression between ob-
servations was determined by dividing the difference between
the scores of each pair of radiographs by the number of months
that had elapsed between them. The progression rate was
expressed as the change in score (total, erosion, or joint space
narrowing) per month; this was annualized when necessary to
express the progression rate per year (36). For each patient
with ⱖ2 radiographic observations, the slope of the least-
squares linear regression line was calculated to estimate the
annualized progression rates of the erosion score, joint space
narrowing score, and total Sharp score, assuming that progres-
sion remains relatively constant over time.
DNA preparations and HLA–DRB1 genotyping.
Genomic DNA was isolated by standard protocol from blood
mononuclear cells obtained from RA patients. Patients were
genotyped using a DNA-based sequence-specific oligonucleo-
tide probe assay for HLA–DRB1*04 alleles (including *0401,
*0404, *0408, *0405, and *0410) as described elsewhere (42).
Confirmatory sequencing or allele-specific quantitative poly-
merase chain reaction (PCR) was used to verify the gene dose
in homozygotes. Distinction between the *0404 and *0408
DRB1*04 alleles was not made because of their structural
similarity and the rarity of allele type *0408. This was also true
for DRB1*04 alleles *0405 and *0410, since *0410 was infre-
quent.
RANKL polymorphism genotyping. The SNP of
RANKL, SNP1 (reference no. rs922996), was determined by
PCR pyrosequencing. The PCR was performed in a 25-␮l
mixture containing 20–40 ng of genomic DNA template,
standard PCR buffer, 200 ␮M dNTPs, 2 mM MgCl2, 200 nM
primers, and 1 unit of AmpliTaq Gold Polymerase (Applied
Biosystems, Foster City, CA). The PCR began with a denatur-
ation step at 95°C for 5 minutes, followed by 50 cycles of
denaturing at 95°C for 15 seconds, annealing at 65°C for 30
seconds, plus extension at 72°C for 15 seconds, and ended with
a final elongation step at 72°C for 5 minutes. Using the forward
primer 5⬘-TTT-CTG-GAC-AGA-GGG-ATT-GG-3⬘ and re-
verse primer 5⬘-TGA-TCT-GGG-AGC-CTC-AAA-AC-3⬘, a
153-bp amplicon was generated and checked for size and
purity by agarose gel electrophoresis. The forward primer was
purified by high-performance liquid chromatography and was
biotinylated at the 5⬘ end to allow immobilization onto
streptavidin-coated Dynabeads M-280 (Dynal Biotech, Oslo,
Norway).
Pyrosequencing of 20 ␮l of the PCR product immobi-
lized onto the beads was performed using the sequencing
primer 5⬘-CAA-CGG-TGG-GGC-A-3⬘ and SNP Reagent kits
(Pyrosequencing, Uppsala, Sweden) according to the manufac-
turer’s instructions. SNP genotype analysis was performed
using the SNP software in a PSQ 96MA System (both from
Pyrosequencing, Uppsala, Sweden).
Genotyping of the RANKL polymorphism (rs5803141;
–/TAAA) was performed by fluorescence-based PCR using the
forward primer 5⬘-TET-TGA-GCT-GAG-ATT-GCA-CCA-
3096 WU ET AL

Table 1. Major baseline characteristics of the 237 patients with early rheumatoid arthritis
Characteristic Value
Age at disease onset, mean ⫾ SD years 50.0 ⫾ 12.9
Female, % 75.5
Ethnicity, %
Caucasian 77.1
Hispanic 15.7
Asian, Pacific Islander 3.4
African American 1.3
Disease duration, mean ⫾ SD months 6.16 ⫾ 3.35
No. of comorbidities per patient, mean ⫾ SD* 1.60 ⫾ 1.66
Patient-reported total annual household income, mean ⫾ SD dollars (⫻1,000) 38.9 ⫾ 26.4
Highest year of education completed, mean ⫾ SD 12.7 ⫾ 2.9
Smoker, % 56.5
Rheumatoid nodules, % 11.5
Extraarticular manifestations other than rheumatoid nodules, %† 9.0
Rheumatoid factor, mean ⫾ SD IU/ml 385 ⫾ 526
Erythrocyte sedimentation rate, mean ⫾ SD mm/hour 42.0 ⫾ 25.4
C-reactive protein, mean ⫾ SD mg/dl 3.12 ⫾ 6.20
Disease Activity Score, mean ⫾ SD 4.71 ⫾ 1.15
Health Assessment Questionnaire disability index, mean ⫾ SD (range 0–3.0) 1.20 ⫾ 0.72
Center for Epidemiologic Studies Depression Scale, mean ⫾ SD (range 0–60) 15.7 ⫾ 11.7

* Frequent comorbidities observed in this cohort were depression, hypertension, broken bones or
fractures, coronary heart disease, gall bladder disease, and osteoporosis. Other comorbidities were cancer
of the lung, breast, uterus, cervix, and prostate, as well as diabetes, hyperthyroidism, cataracts, glaucoma,
myocardial infarction, cardiomyopathy, pneumonia, emphysema, asthma, ulcer, gastritis, endometriosis,
duodenal spasms, back problems, fibrositis, insomnia, allergies, angina, stroke, chronic bronchitis, hiatal
hernia, mitral valve prolapse, diverticulitis, etc.
† Most had self-reported symptoms of keratoconjunctivitis sicca (dry eyes, dry mouth).

CT-3⬘ and the reverse primer 5⬘-TCC-TGA-ATC-TGT-CTC-
CAC-TTC-A-3⬘. The PCR condition and contents were similar
to those described for SNP1 except that the reaction volume
was 5 ␮l. A 207-bp (without the TAAA insertion) or 211-bp
(with the TAAA insertion) amplicon generated by PCR was
mixed with the loading buffer and GeneScan-350 TAMRA size
standard (Applied Biosystems) and subjected to electrophor-
esis in polyacrylamide sequencing gels (Long Ranger Singel
Packs; Cambrex Bioscience, Rockland, ME) for 2 hours at 3
kV using an ABI 377 DNA sequencer (Applied Biosystems).
The gel image was analyzed by GeneScan software version 3.1
for sizing amplified products and Genotyper software version
2.5 for allele assignment (both from Applied Biosystems).
Allele frequencies of RANKL polymorphisms of this cohort
showed no significant departure from expected frequencies
according to Hardy-Weinberg equilibrium.
Statistical analysis. Comparisons between patient
groups were made by 2-sample, 2-sided t-tests for continuous
variables and by chi-square test for categorical variables. The
nonparametric Wilcoxon rank sum test and Fisher’s exact test
were performed when necessary for continuous and categori-
cal variables, respectively. For comparisons of continuous
variables among 3 groups, we used the Kruskal-Wallis non-
parametric test of one-way analysis of variance for ranks. In
addition, to explore the contribution of the DRB1*04 and
RANKL gene polymorphisms to the age at RA onset, we used
a multiple regression model, in which age at onset was
regressed on the presence of the shared epitope, the RANKL
at-risk alleles, and their interaction.
Pairwise linkage disequilibrium between 2 genetic
polymorphisms was assessed using a chi-square test of associ-
ation on phase-known haplotypes from DNA samples. Haplo-
types were inferred using the partition–ligation algorithm to
reconstruct individual haplotypes from population genotype
data (Haplotyper 1.0 Linux software [43]). The extent of
linkage disequilibrium between the 2 RANKL polymorphisms
(D⬘ and r2) was assessed using the GOLD program (44).
Statistical analyses were performed using software packages
SAS release 8.2 (SAS Institute, Cary, NC) and Prism 3.02
(GraphPad Software, San Diego, CA). P values less than 0.05
were considered statistically significant. We considered these
exploratory analyses to be preliminary and hypothesis-
generating; thus, no Bonferroni correction was made.
RESULTS
RA patient characteristics. This study cohort
consisted of 237 patients with early, seropositive, active
RA who had been enrolled within 15 months of symp-
tom onset, before initiation of disease-modifying anti-
rheumatic drug therapy. Table 1 shows the baseline
characteristics of the study patients. Their age at RA
onset was normally distributed, with a mean ⫾ SD of
50.0 ⫾ 12.9 years, and their disease duration at study
entry was 6.16 ⫾ 3.35 months (mean ⫾ SD). Most of the
patients (75.5%) were women and most had active
RANKL AND HLA–DRB1 GENOTYPES AND AGE AT ONSET OF RA 3097

Table 2. Distribution and ethnic variation of the rheumatoid arthritis–associated DRB1*04 alleles*
Total study group DRB1*04⫹ cohort

Caucasians Hispanics Caucasians Hispanics

HLA–DRB1*04 allele All patients (n ⫽ 182) (n ⫽ 37) P (n ⫽ 98) (n ⫽ 17) P
1 copy of a DRB1*04 allele 43 43.5 43.3 NS 100 100 NS
*0401 27 32.4 5.4 0.0005 60.2 11.8 0.0003
*0404 or *0408 12 9.9 29.7 0.0012 18.4 64.7 ⬍0.0001
*0405 or *0410 4 1.1 10.8 0.0082 2.0 23.5 0.0042
2 copies of a DRB1*04 allele 10 10.4 2.7 NS 19.4 5.9 NS
*0401/*0401 homozygous 6 6.0 2.7 NS 11.2 5.9 NS
*0401/*0404 or *0408 heterozygous 2 2.8 0 NS 5.1 0 NS
*0401/*0405 or *0410 heterozygous 1 1.6 0 NS 3.1 0 NS

* Values are percentages. NS ⫽ not significant.

disease (mean DAS 4.71) (45) and moderate functional
disability (mean HAQ disability index 1.20).
Distribution of DRB1*04 alleles. RA-associated
DRB1*04 alleles (DRB1*0401, *0404, *0405, *0408, or
*0410) were present in 53% of the patients; 43% had 1
copy of DRB1*04 (27% had *0401, 12% had *0404 or
*0408, and 4% had *0405 or *0410) and 10% had 2
copies (6% were *0401 homozygotes and 4% had *0401
and other *04 alleles) (Table 2). Comparison of patients
with (n ⫽ 124) and without (n ⫽ 113) RA-associated
DRB1*04 alleles revealed that the presence of the SE
was associated with an earlier age at disease onset
(mean ⫾ SD 47.2 ⫾ 12.7 years versus 53.0 ⫾ 12.5 years
in the SE– group; P ⫽ 0.0004) (Table 3). There was no
trend toward a younger disease onset among patients
with 2 copies of DRB1*04 (50.2 ⫾ 13.6 years; n ⫽ 23)
compared with those with a single copy (46.5 ⫾ 12.4
years; n ⫽ 101) (P ⫽ 0.2).
Ethnic variation in age at disease onset and
RA-associated DRB1*04 alleles. The major ethnic
groups of the cohort were Caucasians (n ⫽ 182 [77%])
and Hispanics (n ⫽ 37 [16%]). The numbers of patients
of other ethnic backgrounds were too small to make a
meaningful comparison. The Hispanic patients had an
earlier disease onset compared with the Caucasian pa-
tients (42.2 ⫾ 11.4 versus 51.6 ⫾ 12.9 years; P ⬍ 0.0001).
This trend was also seen for comparisons of DRB1*04⫹
Hispanics and Caucasians (Table 3). This association
between the SE and an earlier age at RA onset was
observed in Caucasians (48.5 ⫾ 12.4 years in SE⫹ versus
55.3 ⫾ 12.5 years in SE–; P ⫽ 0.003) but not in Hispanics
(40.4 ⫾ 13.1 years in SE⫹ versus 43.7 ⫾ 9.7 years in SE–;
P ⫽ 0.4). The sample size of the Hispanic group was
insufficiently powered to detect a difference between
SE⫹ and SE– subgroups.
The presence of RA-associated DRB1*04 alleles
was higher in the Caucasians (54%) than in the Hispan-
ics (46%). There was a striking difference in the distri-
bution of specific RA-associated DRB1*04 alleles be-
tween the 2 ethnic groups (Table 2). DRB1*0401 was the
most frequent allele in the Caucasians, whereas *0404
(*0408) and *0405 (*0410) accounted for the majority of
SE⫹ Hispanics. Patients who had 2 copies of SE-
containing DRB1*04 represented 10.4% of the Cauca-
sians of this cohort, but only 2.7% of the Hispanics.
HLA–DRB1 and RANKL gene polymorphisms
and age at RA onset. Because RANK and RANKL are
novel therapeutic targets for arthritis, RANKL located
on chromosome 13q14 was chosen as a non-HLA can-
didate gene for RA. We observed that an intronic SNP,

Table 3. Age at onset of rheumatoid arthritis symptoms, according to the presence and absence of DRB1*04 alleles*
Total study group DRB1*04⫹ cohort DRB1*04⫺ cohort P†
All study patients 50.0 ⫾ 12.9 (n ⫽ 237) 47.2 ⫾ 12.7 (n ⫽ 124) 53.0 ⫾ 12.5 (n ⫽ 113) 0.0004
Caucasian patients 51.6 ⫾ 12.9 (n ⫽ 182) 48.5 ⫾ 12.4 (n ⫽ 98) 55.3 ⫾ 12.5 (n ⫽ 84) 0.0003
Hispanic patients 42.2 ⫾ 11.4 (n ⫽ 37) 40.4 ⫾ 13.1 (n ⫽ 17) 43.7 ⫾ 9.7 (n ⫽ 20) NS
P* ⬍0.0001 0.0150 0.0002 –

* The presence of rheumatoid arthritis–associated HLA–DRB1*04 alleles was associated with early onset of the disease. Values
are the mean ⫾ SD years of age.
* For comparisons between Hispanics and Caucasians.
† For comparisons between DRB1*04⫹ and DRB1*04⫺ cohorts. NS ⫽ not significant.
3098 WU ET AL

Table 4. Association between the presence of HLA–DRB1*04 (SE-containing) and RANKL risk alleles and early age at onset
of rheumatoid arthritis*
Total study group (n ⫽ 237) Caucasian patients (n ⫽ 182)

Age at onset, Age at onset,

Genotype mean ⫾ SD years P mean ⫾ SD years P
HLA–DRB1*04
SE⫹ 47 ⫾ 12.7 (n ⫽ 124) 0.0004 48 ⫾ 12.4 (n ⫽ 98) 0.0003
SE⫺ 53 ⫾ 12.5 (n ⫽ 113) 55 ⫾ 12.5 (n ⫽ 84)
RANKL
CC 44 ⫾ 13.5 (n ⫽ 25) 0.0080 44 ⫾ 14.0 (n ⫽ 18) 0.0110
CT plus TT 51 ⫾ 12.7 (n ⫽ 212) 52 ⫾ 12.5 (n ⫽ 164)
RANKL†
TAAA-C⫹/⫹ 44 ⫾ 14.4 (n ⫽ 21) 0.0190 44 ⫾ 14.8 (n ⫽ 16) 0.0140
TAAA-C⫹/⫺ ⫹ TAAA-C⫺/⫺ 51 ⫾ 12.6 (n ⫽ 216) 52 ⫾ 12.5 (n ⫽ 166)
RANKL/DRB1*04
CC/SE⫹ 35 ⫾ 7.2 (n ⫽ 13) 0.0003 36 ⫾ 6.1 (n ⫽ 10) 0.0005
CT plus TT/SE⫹ 49 ⫾ 12.5 (n ⫽ 111) 50 ⫾ 12.1 (n ⫽ 88)
RANKL/DRB1*04
CC/SE⫹ 35 ⫾ 7.2 (n ⫽ 13) 0.0005 36 ⫾ 6.1 (n ⫽ 10) 0.0013
CC/SE⫺ 52 ⫾ 13.2 (n ⫽ 12) 55 ⫾ 14.0 (n ⫽ 8)
RANKL/DRB1*04
CC/SE⫹ 35 ⫾ 7.2 (n ⫽ 13) ⬍0.0001 36 ⫾ 6.1 (n ⫽ 10) ⬍0.0001
CT plus TT/SE⫺ 53 ⫾ 12.5 (n ⫽ 101) 55 ⫾ 12.4 (n ⫽ 76)
RANKL/DRB1*04†
TAAA-C⫹/⫹ and SE⫹ 34 ⫾ 7.2 (n ⫽ 10) ⬍0.0001 35 ⫾ 6.3 (n ⫽ 9) ⬍0.0001
TAAA-C⫹/⫺ ⫹ TAAA-C⫺/⫺ and SE⫺ 53 ⫾ 12.4 (n ⫽ 102) 55 ⫾ 12.3 (n ⫽ 77)

* SE ⫽ shared epitope.
† Haplotype TAAA-C⫹/⫹ ⫽ TAAA-C/TAAA-C, haplotype TAAA-C⫹/⫺ ⫽ TAAA-C/absence of TAAA insertion -T, and
haplotype TAAA-C⫺/⫺ ⫽ homozygous for the absence of TAAA insertion -T.

SNP1 (rs922996; T⬎C), of RANKL was associated with
age at RA symptom onset. As shown in Table 4, among
the 237 study patients, those with RANKL-CC genotypes
had earlier ages at disease onset (44 ⫾ 13.5 years) than
did those with RANKL-CT/TT genotypes (51 ⫾ 12.7
years) (P ⫽ 0.0080). The presence of RA-associated
DRB1*04 alleles was associated with an earlier disease
onset (47 ⫾ 12.7 years versus 53 ⫾ 12.5 years in SE–
patients; P ⫽ 0.0004). In the presence of SE-containing
DRB1*04 alleles, RA patients with RANKL-CC geno-
types had an earlier age at RA onset than did those with
CT/TT genotypes (35 ⫾ 7.2 years versus 49 ⫾ 12.5 years;
P ⫽ 0.0003). In the absence of SE-containing DRB1*04
alleles, the RANKL-CC genotype was not associated
with an early age at RA onset compared with the CT/TT
genotypes (52 ⫾ 13.2 years versus 53 ⫾ 12.5 years).
The mean age at disease onset in RA patients
harboring SE plus RANKL-CC (35 ⫾ 7.2 years) was an
average of 18 years younger than that in patients without
SE but with RANKL-CT/TT (53 ⫾ 12.5 years; P ⬍
0.0001). Similarly, among the 182 Caucasian RA pa-
tients, those who had SE plus RANKL-CC (36 ⫾ 6.1
years) developed RA an average of 19 years earlier than
did those without the RA-associated DRB1*04 alleles
who had RANKL-CT/TT (55 ⫾ 12.4 years; P ⬍ 0.0001)
(Table 4).
A multiple regression model, in which age at RA
onset was regressed on the presence of SE, RANKL
at-risk allele, and their interaction, showed that age at
onset was decreased an average of 18 years in the total
group of patients with both SE and RANKL-CC alleles
in comparison with patients with neither allele (age at
onset ⫽ 53.10 – 4.6 ⫻ SE – 0.8 ⫻ RANKL-CC – 12.5
[SE ⫻ RANKL-CC] interaction). Individual P values
were P ⬍ 0.007 for the presence of SE, P ⫽ 0.85 for the
presence of RANKL-CC, P ⬍ 0.02 for the interaction of
these 2 genotypes, and P ⬍ 0.0001 for the model. Similar
multiple regression modeling resulted in an average 19.6
years younger age at RA onset in Caucasian patients
with both SE and RANKL-CC alleles in comparison with
patients with neither allele (age at onset ⫽ 55.37 – 5.5 ⫻
SE – 0.4 ⫻ RANKL-CC – 13.7 [SE ⫻ RANKL-CC]
interaction; P ⬍ 0.0001 for the model). This further
supported the idea that an interaction between SE-
containing DRB1*04 and RANKL-CC genotypes might
contribute to the younger age at RA onset. These P
values remained significant after correction for multiple
testing of 3 gene polymorphisms (P ⬍ 0.0003). The
RANKL AND HLA–DRB1 GENOTYPES AND AGE AT ONSET OF RA
Figure 1. Proportion of patients with the rheumatoid arthritis (RA)–
associated DRB1*04 and RANKL-CC alleles, according to age at RA
onset (3rd decade ⫽ ages 20–29 years; 4th decade ⫽ ages 30–39 years;
5th decade ⫽ ages 40–49 years; 6th decade ⫽ ages 50–59 years; 7th
decade ⫽ 60–69 years; 8th decade ⫽ 70–79 years).
combination of both at-risk alleles lowered the age at
disease onset much more than did each individual risk
genotype, but the age effect of the RANKL-CC genotype
was not observed in the absence of SE (Table 4).
We grouped these 237 RA patients according to
their age at RA onset (by decades) to assess whether the
proportion of patients harboring the RA-associated
DRB1*04 alleles, the RANKL-CC genotype, and
DRB1*04 plus RANKL-CC genotypes varied according
to age at disease onset (Figure 1). Strikingly, 12 of the 13
RA patients who had an onset between 20 and 30 years
of age had at least 1 copy of the RA-associated DRB1*04
alleles (Figure 1A). The majority of the patients who
developed RA before age 50 (74 of 119 [62%]) had at
3099
least 1 copy of the RA-associated DRB1*04 alleles, while
the majority of the patients who developed RA after age
50 (68 of 118 [58%]) lacked these DRB1*04 alleles
(Figure 1A).
Although only a few patients had RANKL-CC, a
higher proportion of those who had the RANKL-CC
genotype developed the disease before age 40 (Figure
1B). The co-occurrence of RA-associated DRB1*04 and
RANKL-CC genotypes was present in 23% of patients
who developed the disease during the third decade, 16%
who developed it during the fourth decade, 5% who
developed it fifth decade, and was absent in patients who
developed RA after the age of 50 years (Figure 1C). It
appeared that the co-occurrence of RA-associated
DRB1*04 and RANKL-CC genotypes was more preva-
lent in a subset of patients who developed RA at a
younger age, but the majority of our early RA patients
did not have these genotypes. Our results suggested that
1) the co-occurrence of DRB1*04 and RANKL-CC ge-
notypes increased the disease penetrance, resulting in an
earlier age at onset of RA, and 2) there might be genetic
differences between RA patients with younger age at
onset and those with older age at onset.
The tested SNP1 was predicted by the TFSearch
database to affect binding of transcription factors, which
might cause differential gene expression of RANKL.
SNP1 located in the 3⬘ end of the gene (Figure 2) might
be in linkage disequilibrium with the downstream gene.
To confirm the expression of RANKL rather than its
downstream gene, we tested another intronic polymor-
phism (the presence or absence of the 4-nucleotide
TAAA insertion; –/TAAA, rs5803141) in the 5⬘ region
of RANKL located 37.5 kb upstream of SNP1 (Figure 2).
Significant linkage disequilibrium between these 2 poly-
morphisms was observed in our Caucasian patients
(D⬘ ⫽ 0.734, r2 ⫽ 0.52, P ⬍ 0.000001).
Figure 2. The RANKL gene structure and the relative positions of
polymorphisms tested in this study. This gene located on 13q14 covers
45.28 kb, from 40,934,872 to 40,980,153 bp (National Center for
Biotechnology Information build 34; August 2003). The positions of
exons 1–8 are marked; exons 6 and 7 are physically close and are
indistinguishable on the scale used here. The SNP1 (T⬎C at
40,976,671 bp) and the insertion polymorphism (absence or presence
of the 4-nucleotide TAAA insertion; –/TAAA, at 40,939,161 bp) are
37.5 kb apart.
3100 WU ET AL

Table 5. Lack of association between the RANKL SNP1 genotype and the yearly radiographic progression rate*
Erosions, Joint space narrowing, Total progression,
median (IQR) median (IQR) median (IQR)
Total (n ⫽ 190)
CC (n ⫽ 20) 0.73 (⫺0.0068, 3.17) 0.17 (⫺0.45, 1.69) 0.97 (⫺0.39, 4.77)
CT (n ⫽ 112) 0.48 (⫺0.041, 1.82) 0.025 (⫺0.11, 1.23) 1.01 (⫺0.12, 2.74)
TT (n ⫽ 58) 0.68 (0, 1.55) 0.33 (⫺0.075, 1.30) 1.14 (⫺0.037, 2.89)
CT ⫹ TT (n ⫽ 170) 0.49 (⫺0.023, 1.79) 0.081 (⫺0.11, 1.26) 1.06 (⫺0.10, 2.87)
P (CC vs. CT vs. TT)† 0.62 0.76 0.90
P (CC vs. [CT ⫹ TT])‡ 0.33 0.73 0.94
Caucasian patients (n ⫽ 149)
CC (n ⫽ 13) 0.55 (⫺0.027, 1.5) 0.064 (⫺1.17, 1.33) 0.82 (⫺1.30, 4.5)
CT (n ⫽ 89) 0.40 (⫺0.053, 1.79) 0.040 (⫺0.13, 1.07) 0.97 (⫺0.20, 2.48)
TT (n ⫽ 47) 0.66 (0, 1.49) 0.46 (⫺0.070, 1.25) 1.16 (⫺0.069, 2.87)
CT ⫹ TT (n ⫽ 136) 0.45 (⫺0.043, 1.62) 0.081 (⫺0.11, 1.09) 1.06 (⫺0.16, 2.59)
P (CC vs. CT vs. TT)† 0.94 0.42 0.61
P (CC vs. [CT ⫹ TT])‡ 0.73 0.32 0.47

* Values are the total Sharp score units per year. IQR ⫽ interquartile range (25th, 75th percentiles) of radiographic
progression.
† By Kruskal-Wallis nonparametric test for one-way analysis of variance.
‡ By Wilcoxon’s nonparametric rank-sum test.

These 2 polymorphisms were used to construct
haplotypes, which showed the absence of insertion (-)-T
(54.43%), TAAA-C (33.12%), TAAA-T (6.33%), and
--C (6.12%) in our cohort. The frequencies of these 4
haplotypes were 54.1%, 33.1%, 6.6%, and 5.8% in the
Caucasians and 56.6%, 31.6%, 6.6%, and 5.3% in the
Hispanics, showing no significant difference between the
2 ethnic groups (P ⫽ 0.98). RA patients with 2 copies of
the TAAA-C haplotype had a younger age at disease
onset (44 ⫾ 14.4 years) compared with patients with 1
copy or with no copy of the TAAA-C haplotype (51 ⫾
12.6 years; P ⫽ 0.019) (Table 4). The presence of
RANKL TAAA-C⫹/⫹ and SE⫹ was associated with
younger age at RA onset (34 ⫾ 7.2 years) compared with
the presence of RANKL TAAA-C⫹/– plus TAAA-C–/–
and SE– (53 ⫾ 12.4 years; P ⬍ 0.0001) (Table 4). These
data supported the idea that RANKL polymorphisms
and SE containing HLA–DRB1*04 alleles might interact
to predispose an individual to the early onset of RA.
RANKL gene polymorphisms and erosive disease.
The role of RANK/RANKL in the regulation of bone
destruction provided a rationale for assessing the asso-
ciation of RANKL genotypes with the presence of ero-
sive disease at baseline and with the radiographic rate of
progression. At study entry, 63 of 217 patients (29%)
had erosive damage, as indicated by radiographic ero-
sion scores ⱖ2 (20 patients had no radiographic data at
study entry). Six of these 63 patients had the
RANKL-CC genotype. Among the 154 patients with
radiographic erosion scores ⬍2, 17 had the RANKL-CC
genotype. No significant association between the
RANKL genotype and the erosion score at baseline was
observed (P ⫽ 0.7). There was no statistically significant
difference in the yearly radiographic progression rate
(erosion, joint space narrowing, and total Sharp scores)
in patients with the RANKL-CC genotype compared
with patients with the RANKL-CT/TT genotypes among
the 149 Caucasians and the 190 total group patients in
whom serial radiographs were obtained (Table 5). Sim-
ilarly, no significant association between the RANKL
haplotype and the yearly radiographic progression rate
was observed in this cohort (data not shown).
DISCUSSION
This study explored one well-known genetic fac-
tor for RA (HLA–DRB1*0401, *0404, *0405, *0408
alleles) and one novel candidate gene (RANKL) for their
potential roles in the development of RA. Using a
cohort of patients with early RA, we confirmed the
report of (8–10) an association between DRB1*04 alleles
and ages at RA onset (P ⫽ 0.0004) (Table 3) and the
ethnic variation (36) in the RA-associated DRB1*04
alleles (Table 2). We observed an average of 6 years
younger age at RA onset among patients with SE-
containing DRB1*04 alleles. The novel RANKL poly-
morphisms (the insertion of TAAA, SNP1-C, and the
TAAA-C haplotype) were also associated with an earlier
age at RA onset (⬃7 years). Most interesting was our
observation that the co-occurrence of DRB1*04 and
RANKL risk alleles (haplotypes) was associated with a
much younger age at RA onset, an effect of ⬃18–20
RANKL AND HLA–DRB1 GENOTYPES AND AGE AT ONSET OF RA
years (Table 4). The co-occurrence of DRB1*04 and
RANKL risk alleles appeared to have a stronger impact
on the earlier age at disease onset than did each gene
variant individually. Of note, this effect of RANKL risk
alleles was present only among patients who had SE-
containing DRB1*04 alleles. These RANKL genotypes
appeared not to be associated with erosive disease at
baseline or with the yearly rate of progression of radio-
graphic joint damage (Table 5).
We observed that the majority of the patients
who developed RA symptoms in their third to fifth
decades of life carried at least 1 copy of the DRB1*04
alleles. In contrast, fewer than half of the patients who
developed RA in their sixth or later decades had these
alleles (Figure 1A). Interestingly, 92% of the patients
who were diagnosed as having RA between the ages of
20 and 30 years carried at least 1 of the RA-associated
DRB1*04 alleles, suggesting strong influences of the SE
in the early onset of RA.
The most prevalent DRB1*04 allele in this cohort
was *0401. Genotypes containing DRB1*0401, unlike
other SE-containing DRB1 alleles, have been shown to
increase susceptibility or penetrance even if only 1 copy
is present (8). This might explain our failure to find an
SE dose effect on the age at onset. To our knowledge,
there has been no clear demonstration of an SE dose
effect on the age at RA onset (8). The proportion of RA
patients harboring both DRB1*04 and RANKL risk
alleles was highest in patients who developed the disease
during their third decade of life (23%), exhibited a
declining trend among patients diagnosed during their
fourth (16%) and fifth (5%) decades, and was absent in
patients diagnosed after age 50 years (Figure 1C). These
results suggested that genetic interactions between
HLA–DRB1*04 and RANKL genotypes enhanced dis-
ease penetrance and provided evidence for genetic
differences between early-onset and late-onset RA pa-
tients. We did not genotype other, non-DRB1*04 SE
alleles (*01, *10, *1402), since the presence of these
alleles has not been reported to confer an increase in
either susceptibility or disease severity of seropositive
RA among North American Caucasians (8,46–48).
The age at onset of RA symptoms in Hispanics
was earlier compared with that in Caucasians (42.2 ⫾
11.4 years versus 51.6 ⫾ 12.9 years; P ⬍ 0.0001) (Table
3). This ⬃8-year difference was also observed among
SE⫹ Hispanics compared with SE⫹ Caucasians (40.4 ⫾
13.1 years versus 48.5 ⫾ 12.4 years; P ⫽ 0.015) (Table 3).
Compared with Caucasian patients, Hispanic patients
showed ethnic variations in the frequency and distribu-
tion of the specific SE-containing DRB1*04 alleles (Ta-
3101
ble 2), similar to previous findings by Del Rincon et al
(49). While the Hispanic patients in our study and the
study by Del Rincon were mainly of Mexican descent,
Del Rincon and colleagues found no difference in the
age at symptom onset (or age at diagnosis) between their
Hispanic and Caucasian RA patients (49). We have no
explanation for this difference except for the possible
difference in inclusion criteria: one of the inclusion criteria
for our cohort but not theirs was rheumatoid factor sero-
positivity.
We observed strong linkage disequilibrium (P ⬍
0.0001) between the 2 RANKL polymorphisms (the
insertion polymorphism and SNP1 shown in Figure 2)
that are 37.5 kb apart within the entire 45.3-kb gene.
Either of the polymorphisms or the combined haplotype
was associated with an earlier age at RA onset, suggest-
ing a role of RANKL variants in the pathogenesis of RA.
The interaction between RANKL, RANK, and OPG is
pivotal in bone and cartilage homeostasis, as well as in
the regulation of bone metabolism and immune func-
tions (23,28). The RANKL polymorphisms we tested, or
other polymorphisms in linkage disequilibrium, might
affect levels of gene expression, resulting in differential
expression of RANKL protein. It is tempting to specu-
late that an abnormal level of RANKL expression might
dysregulate communication between dendritic cells and
T cells, resulting in a breakdown of the immune toler-
ance to self components in the joint, while the presence
of SE-containing HLA–DRB1 alleles might shape the T
cell repertoire, leading to enhanced T cell reactivity
toward arthritogenic antigens. The co-occurrence of
DRB1 and RANKL at-risk genotypes might accelerate
escape from tolerance, resulting in a much earlier devel-
opment of arthritis.
In summary, the results of our study provide the
first evidence of a strong association between an early
age at onset of RA and the combined presence of
HLA–DRB1 and RANKL genotypes. However, in view
of the large number of genetic polymorphisms that were
not tested by us, it is possible that our findings could
have occurred by chance and, at this time, should be
considered preliminary. Further studies are needed to
confirm this association and to elucidate the role attrib-
uted to RANKL polymorphisms in the pathogenesis of
RA.
ACKNOWLEDGMENTS
We thank all of the participating patients, and we
thank Elly Park for technical help in this study.
3102
REFERENCES
1. Jawaheer D, Gregersen PK. Rheumatoid arthritis: the genetic
components. Rheum Dis Clin North Am 2002;28:1–15.
2. Huizinga TW. Genetics in rheumatoid arthritis. Best Pract Res
Clin Rheumatol 2003;17:703–16.
3. Cornelis F, Faure S, Martinez M, Prud’homme JF, Fritz P, Dib C,
et al. New susceptibility locus for rheumatoid arthritis suggested by
a genome-wide linkage study. Proc Natl Acad Sci U S A 1998;95:
10746–50.
4. Jawaheer D, Seldin MF, Amos CI, Chen WV, Shigeta R, Monteiro
J, et al. A genomewide screen in multiplex rheumatoid arthritis
families suggests genetic overlap with other autoimmune diseases.
Am J Hum Genet 2001;68:927–36.
5. Jawaheer D, Seldin MF, Amos CI, Chen WV, Shigeta R, Etzel C,
et al. Screening the genome for rheumatoid arthritis susceptibility
genes: a replication study and combined analysis of 512 multicase
families. Arthritis Rheum 2003;48:906–16.
6. MacKay K, Eyre S, Myerscough A, Milicic A, Barton A, Laval S,
et al. Whole-genome linkage analysis of rheumatoid arthritis
susceptibility loci in 252 affected sibling pairs in the United
Kingdom. Arthritis Rheum 2002;46:632–9.
7. Gregersen PK, Silver J, Winchester RJ. The shared epitope
hypothesis: an approach to understanding the molecular genetics
of susceptibility to rheumatoid arthritis. Arthritis Rheum 1987;30:
1205–13.
8. Fries JF, Wolfe F, Apple R, Erlich H, Bugawan T, Holmes T, et al.
HLA–DRB1 genotype associations in 793 white patients from a
rheumatoid arthritis inception cohort: frequency, severity, and
treatment bias. Arthritis Rheum 2002;46:2320–9.
9. MacGregor A, Ollier W, Thomson W, Jawaheer D, Silman A.
HLA-DRB1*0401/0404 genotype and rheumatoid arthritis: in-
creased association in men, young age at onset, and disease
severity. J Rheumatol 1995;22:1032–6.
10. Gonzalez-Gay MA, Garcia-Porrua C, Hajeer AH. Influence of
human leukocyte antigen-DRB1 on the susceptibility and severity
of rheumatoid arthritis. Semin Arthritis Rheum 2002;31:355–60.
11. Massardo L, Gareca N, Cartes MA, Cervilla V, Gonzalez A,
Jacobelli S. The presence of the HLA-DRB1 shared epitope
correlates with erosive disease in Chilean patients with rheumatoid
arthritis. Rheumatology (Oxford) 2002;41:153–6.
12. Mattey DL, Dawes PT, Gonzalez-Gay MA, Garcia-Porrua C,
Thomson W, Hajeer AH, et al. HLA-DRB1 alleles encoding an
aspartic acid at position 70 protect against development of rheu-
matoid arthritis. J Rheumatol 2001;28:232–9.
13. Mattey DL, Hassell AB, Dawes PT, Cheung NT, Poulton KV,
Thomson W, et al. Independent association of rheumatoid factor
and the HLA–DRB1 shared epitope with radiographic outcome in
rheumatoid arthritis. Arthritis Rheum 2001;44:1529–33.
14. Plant MJ, Jones PW, Saklatvala J, Ollier WE, Dawes PT. Patterns
of radiological progression in early rheumatoid arthritis: results of
an 8 year prospective study. J Rheumatol 1998;25:417–26.
15. Wagner U, Kaltenhauser S, Sauer H, Arnold S, Seidel W,
Hantzschel H, et al. HLA markers and prediction of clinical course
and outcome in rheumatoid arthritis. Arthritis Rheum 1997;40:
341–51.
16. Gough A, Faint J, Salmon M, Hassell A, Wordsworth P, Pilling D,
et al. Genetic typing of patients with inflammatory arthritis at
presentation can be used to predict outcome. Arthritis Rheum
1994;37:1166–70.
17. Emery P, Salmon M, Bradley H, Wordsworth P, Tunn E, Bacon
PA, et al. Genetically determined factors as predictors of radio-
logical change in patients with early symmetrical arthritis. BMJ
1992;305:1387–9.
18. Reveille JD, Alarcon GS, Fowler SE, Pillemer SR, Neuner R,
Clegg DO, et al, for the MIRA Trial Group. HLA–DRB1 genes
WU ET AL
and disease severity in rheumatoid arthritis. Arthritis Rheum
1996;39:1802–7.
19. Weyand CM, Hicok KC, Conn DL, Goronzy JJ. The influence of
HLA-DRB1 genes on disease severity in rheumatoid arthritis. Ann
Intern Med 1992;117:801–6.
20. Weyand CM, Goronzy JJ. Disease mechanisms in rheumatoid
arthritis: gene dosage effect of HLA–DR haplotypes. J Lab Clin
Med 1994;124:335–8.
21. Seldin MF, Amos CI, Ward R, Gregersen PK. The genetics
revolution and the assault on rheumatoid arthritis. Arthritis
Rheum 1999;42:1071–9.
22. Firestein GS. Evolving concepts of rheumatoid arthritis. Nature
2003;423:356–61.
23. Hofbauer LC, Heufelder AE. The role of osteoprotegerin and
receptor activator of nuclear factor ␬B ligand in the pathogenesis
and treatment of rheumatoid arthritis. Arthritis Rheum 2001;44:
253–9.
24. Kong YY, Yoshida H, Sarosi I, Tan HL, Timms E, Capparelli C,
et al. OPGL is a key regulator of osteoclastogenesis, lymphocyte
development and lymph-node organogenesis. Nature 1999;397:
315–23.
25. Ziolkowska M, Kurowska M, Radzikowska A, Luszczykiewicz G,
Wiland P, Dziewczopolski W, et al. High levels of osteoprotegerin
and soluble receptor activator of nuclear factor ␬B ligand in serum
of rheumatoid arthritis patients and their normalization after
anti–tumor necrosis factor ␣ treatment. Arthritis Rheum 2002;46:
1744–53.
26. Kong YY, Feige U, Sarosi I, Bolon B, Tafuri A, Morony S, et al.
Activated T cells regulate bone loss and joint destruction in
adjuvant arthritis through osteoprotegerin ligand. Nature 1999;
402:304–9.
27. Pettit AR, Ji H, von Stechow D, Muller R, Goldring SR, Choi Y,
et al. TRANCE/RANKL knockout mice are protected from bone
erosion in a serum transfer model of arthritis. Am J Pathol
2001;159:1689–99.
28. Nakashima T, Wada T, Penninger JM. RANKL and RANK as
novel therapeutic targets for arthritis. Curr Opin Rheumatol
2003;15:280–7.
29. Alarcon-Riquelme ME. A RUNX trio with a taste for autoimmu-
nity. Nat Genet 2003;35:299–300.
30. Helms C, Cao L, Krueger JG, Wijsman EM, Chamian F, Gordon
D, et al. A putative RUNX1 binding site variant between
SLC9A3R1 and NAT9 is associated with susceptibility to psoriasis.
Nat Genet 2003;35:349–56.
31. Prokunina L, Castillejo-Lopez C, Oberg F, Gunnarsson I, Berg L,
Magnusson V, et al. A regulatory polymorphism in PDCD1 is
associated with susceptibility to systemic lupus erythematosus in
humans. Nat Genet 2002;32:666–9.
32. Tokuhiro S, Yamada R, Chang X, Suzuki A, Kochi Y, Sawada T,
et al. An intronic SNP in a RUNX1 binding site of SLC22A4,
encoding an organic cation transporter, is associated with rheu-
matoid arthritis. Nat Genet 2003;35:341–8.
33. Heinemeyer T, Wingender E, Reuter I, Hermjakob H, Kel
AE, Kel OV, et al. Databases on transcriptional regulation:
TRANSFAC, TRRD and COMPEL. Nucleic Acids Res 1998;26:
362–7.
34. Paulus HE, Bulpitt KJ, Ramos B, Park G, Wong WK. Relative
contributions of the components of the American College of
Rheumatology 20% criteria for improvement to responder status
in patients with early seropositive rheumatoid arthritis. Arthritis
Rheum 2000;43:2743–50.
35. Paulus HE, Wiesner J, Bulpitt KJ, Patnaik M, Law J, Park GS, et
al. Autoantibodies in early seropositive rheumatoid arthritis, be-
fore and during disease modifying antirheumatic drug treatment.
J Rheumatol 2002;29:2513–20.
36. Paulus HE, Oh M, Sharp JT, Gold RH, Wong WK, Park GS, et al.
Correlation of single time-point damage scores with observed
RANKL AND HLA–DRB1 GENOTYPES AND AGE AT ONSET OF RA
progression of radiographic damage during the first 6 years of
rheumatoid arthritis. J Rheumatol 2003;30:705–13.
37. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF,
Cooper NS, et al. The American Rheumatism Association 1987
revised criteria for the classification of rheumatoid arthritis.
Arthritis Rheum 1988;31:315–24.
38. Van der Heijde DM, van ’t Hof MA, van Riel PL, Theunisse LM,
Lubberts EW, van Leeuwen MA, et al. Judging disease activity in
clinical practice in rheumatoid arthritis: first step in the develop-
ment of a disease activity score. Ann Rheum Dis 1990;49:916–20.
39. Blalock SJ, DeVellis RF, Brown GK, Wallston KA. Validity of the
Center for Epidemiological Studies Depression Scale in arthritis
populations. Arthritis Rheum 1989;32:991–7.
40. Sharp JT, Lidsky MD, Collins LC, Moreland J. Methods of scoring
the progression of radiologic changes in rheumatoid arthritis:
correlation of radiologic, clinical and laboratory abnormalities.
Arthritis Rheum 1971;14:706–20.
41. Sharp JT, Young DY, Bluhm GB, Brook A, Brower AC, Corbett
M, et al. How many joints in the hands and wrists should be
included in a score of radiologic abnormalities used to assess
rheumatoid arthritis? Arthritis Rheum 1985;28:1326–35.
42. O’Dell JR, Nepom BS, Haire C, Gersuk VH, Gaur L, Moore GF,
et al. HLA–DRB1 typing in rheumatoid arthritis: predicting
response to specific treatments. Ann Rheum Dis 1998;57:209–13.
43. Niu T, Qin ZS, Xu X, Liu JS. Bayesian haplotype inference for
3103
multiple linked single-nucleotide polymorphisms. Am J Hum
Genet 2002;70:157–69.
44. Abecasis GR, Cookson WO. GOLD—graphical overview of link-
age disequilibrium. Bioinformatics 2000;16:182–3.
45. Van Gestel AM, Prevoo ML, van ’t Hof MA, van Rijswijk MH, van
de Putte LB, van Riel PL. Development and validation of the
European League Against Rheumatism response criteria for rheu-
matoid arthritis: comparison with the preliminary American Col-
lege of Rheumatology and the World Health Organization/Inter-
national League Against Rheumatism criteria. Arthritis Rheum
1996;39:34–40.
46. Nepom GT, Gersuk V, Nepom BS. Prognostic implications of
HLA genotyping in the early assessment of patients with rheuma-
toid arthritis. J Rheumatol Suppl 1996;44:5–9.
47. Nepom GT, Nepom BS. Prediction of susceptibility to rheumatoid
arthritis by human leukocyte antigen genotyping. Rheum Dis Clin
North Am 1992;18:785–92.
48. Nepom GT, Byers P, Seyfried C, Healey LA, Wilske KR, Stage D,
et al. HLA genes associated with rheumatoid arthritis: identifica-
tion of susceptibility alleles using specific oligonucleotide probes.
Arthritis Rheum 1989;32:15–21.
49. Del Rincon ID, I, Battafarano DF, Arroyo RA, Murphy FT,
Fischbach M, Escalante A. Ethnic variation in the clinical mani-
festations of rheumatoid arthritis: role of HLA–DRB1 alleles.
Arthritis Rheum 2003;49:200–8.