ARTHRITIS & RHEUMATISM
Vol. 65, No. 1, January 2013, pp 186–196
© 2013, American College of Rheumatology
A Novel Human Autoantigen, Endothelial Cell Growth Factor,
Is a Target of T and B Cell Responses in
Patients With Lyme Disease
Elise E. Drouin,1Robert J. Seward,1Klemen Strle,1Gail McHugh,1Kianoosh Katchar,1
Diana London ˜o,1Chunxiang Yao,2Catherine E. Costello,2and Allen C. Steere1
Objective. Autoantigen presentation by HLA–DR
molecules is thought to be a central component of many
autoimmune diseases, but identifying disease-relevant
autoantigens has been a difficult challenge. In this study
we aimed to identify autoantigens in patients with
antibiotic-refractory Lyme arthritis, in which infection-
induced autoimmunity is thought to play an important
Methods. Using tandem mass spectrometry, nat-
urally presented HLA–DR self peptides from a patient’s
synovium were identified, synthesized, and reacted with
his peripheral blood mononuclear cells (PBMCs). Im-
munoreactive peptides and their source proteins were
then tested for T and B cell responses using large
numbers of patient cells or sera.
Results. Of 120 HLA–DR–presented self peptides
identified from one patient, one peptide derived from
endothelial cell growth factor (ECGF) caused his
PBMCs to proliferate. T and B cell responses to ECGF
occurred systemically in ?10–30% of patients with early
or late manifestations of Lyme disease, primarily in
those with refractory arthritis–associated HLA–DR al-
leles, such as DRB1*0101 and 0401. Compared with
patients with antibiotic-responsive arthritis, those with
antibiotic-refractory arthritis had significantly higher
concentrations of ECGF in synovial fluid (P < 0.0001)
and more often had ECGF antibody reactivity. Among
non–antibiotic-treated historical patients who devel-
oped arthritis, 26% had ECGF reactivity, which often
developed before the onset of arthritis and was associ-
ated with significantly longer courses of arthritis.
Conclusion. T and B cell responses to ECGF
occur in a subset of patients with Lyme disease, partic-
ularly in those with antibiotic-refractory arthritis, pro-
viding the first direct evidence of autoimmune T and B
cell responses in this illness.
Presentation of autoantigens by HLA–DR mole-
cules to CD4? T cells is thought to be a central com-
ponent of many autoimmune diseases (1). Despite the
strong genetic correlation between HLA–DR alleles and
autoimmunity (2), disease-relevant autoantigens pre-
sented by HLA–DR molecules have often remained
elusive. Furthermore, in autoimmune diseases such as
rheumatoid arthritis (RA) or lupus, multiple autoanti-
gens are thought to be involved, and autoantibodies are
often present months or years before the onset of
clinical disease (3,4), suggesting that additional critical
factors are required to trigger tissue pathology (3). Even
so, recognition of self antigens is an essential component
in the development of disease pathology.
Lyme arthritis, a late manifestation of infection
Supported by the NIH (grants AR-20358 [to Dr. Steere],
P41-GM-104603/RR-10888, S10-RR-15942, and S10-RR-20946, and
contracts N01-HV-28178 and N01-HV-00239 [to Dr. Costello]); the
Dana Foundation (grants to Drs. Costello and Steere); the Mathers
Foundation; the English, Bonter, Mitchell Foundation; the Eshe Fund;
and the Lyme/Arthritis Research Fund at Massachusetts General
Hospital (grant to Dr. Steere). Dr. Strle is recipient of postdoctoral
fellowships from the Arthritis Foundation and the Walter J. and
Lille A. Berbecker Foundation. Dr. Katchar is recipient of a scholar-
ship for the study of Lyme disease from the Lillian B. Davey
1Elise E. Drouin, PhD, Robert J. Seward, PhD, Klemen Strle,
PhD, Gail McHugh, MS, Kianoosh Katchar, PhD, Diana London ˜o,
MD, Allen C. Steere, MD: Massachusetts General Hospital and
Harvard Medical School, Boston, Massachusetts;
PhD, Catherine E. Costello, PhD: Boston University School of Med-
icine, Boston, Massachusetts.
Drs. Drouin, Seward, Costello, and Steere have a patent
application pending for endothelial cell growth factor antibody testing.
Dr. Steere has received consulting fees, speaking fees, and/or hono-
raria from Merck and Alere (less than $10,000 each).
Address correspondence to Elise E. Drouin, PhD, Massachu-
setts General Hospital, CNY 149/8301, 55 Fruit Street, Boston, MA
02114. E-mail: firstname.lastname@example.org.
Submitted for publication May 28, 2012; accepted in revised
form September 27, 2012.
with the tick-borne spirochete Borrelia burgdorferi (Bb)
(5,6), provides an important human model to study
questions surrounding infection-induced autoimmunity.
Lyme arthritis can usually be treated successfully
with 1–2 months of oral or intravenous (IV) antibiotics;
the successfully treated condition is called antibiotic-
responsive arthritis (7). However, in a small percentage
of patients, proliferative synovitis persists for months
or several years after apparent spirochetal killing with
?3 months of oral and IV antibiotics; this is referred to
as antibiotic-refractory arthritis (8).
This disease course has been postulated to result
from either persistent infection, retained spirochetal
antigens, or infection-induced autoimmunity (9,10). As
evidence against the persistent infection hypothesis,
polymerase chain reaction and culture results of syno-
vectomy specimens obtained in the postantibiotic period
have been uniformly negative (11), and relapse of infec-
tion has not been observed with the use of disease-
modifying antirheumatic drugs (DMARDs) after anti-
biotic therapy (8). Contrary to what might be expected
with retained spirochetal antigens, T and B cell re-
sponses to Bb decline similarly in patients with refrac-
tory arthritis and those with responsive arthritis (12,13),
whereas levels of inflammatory mediators in synovial
fluid (SF), particularly interferon-? (IFN?), remain high
or even increase in patients with refractory arthritis
during the postantibiotic period (14). In support of the
autoimmunity hypothesis, specific HLA–DR alleles,
particularly the DRB1*0101 or 0401 alleles, are the
greatest known genetic risk factor for antibiotic-
refractory arthritis (15). As in other chronic inflamma-
tory arthritides, HLA–DR molecules in antibiotic-
refractory Lyme arthritis are intensely expressed in
inflamed synovium (16).
In a search for molecular mimicry between spi-
rochete and host proteins, partial sequence homology
was found between the human peptides, LFA-1?L332–340
(17) and MAWD-BP280–288(18), and an epitope of Bb
outer surface protein A (OspA163–175) (19), which binds
refractory arthritis–associated HLA–DR molecules (15).
However, only a minority of patients had low-level T cell
reactivity with these self peptides, and none had auto-
antibody responses to these self proteins (17,18,20). In a
later study, Ghosh et al identified human cytokeratin 10
as a cross-reactive target ligand recognized by anti-OspA
antibodies in a small group of patients with refractory
arthritis (3 of 15), but not in those with responsive
arthritis (0 of 5) (21). Finally, several neural proteins
have been reported to induce T or B cell responses in
patients with neuroborreliosis (22–24) or post-Lyme
syndrome (25). However, responses against neural pro-
teins would be unlikely to explain antibiotic-refractory
In this study, we used discovery-based proteomics
and translational research in an effort to identify auto-
antigens in synovial tissue, the target tissue of the
immune attack in antibiotic-refractory Lyme arthritis.
Based on this approach, we report here the identifica-
tion of a novel autoantigen, endothelial cell growth
factor (ECGF), as the target of T and B cell responses in
a subset of patients with Lyme disease, thereby providing
the first direct evidence of autoimmune T and B cell
responses in this illness.
PATIENTS AND METHODS
Patients. All patients with Lyme disease met the Cen-
ters for Disease Control and Prevention criteria (26) and those
with RA met the 2010 American College of Rheumatology/
European League Against Rheumatism criteria (27). Studies
conducted from 1975 to 1987 were approved by the Human
Investigations Committee at the Yale University School of
Medicine (New Haven, CT), those conducted from 1988 to
2002 were approved by Tufts Medical Center (Boston, MA),
and those conducted after 2002 were approved by Massachu-
setts General Hospital.
For patients with erythema migrans (EM), who were
seen from 1998 through 2001 in a study of early Lyme disease,
all available peripheral blood mononuclear cell (PBMC) and
serum samples from culture-positive patients were tested. For
antibiotic-treated patients with Lyme arthritis, who were seen
from 1988 through 2010 in a study called Immunity in Lyme
Arthritis, the first available sample was used. For comparison,
serial serum samples from non–antibiotic-treated patients
who were seen in the late 1970s were tested. For patients with
RA, who were evaluated from 2008 through 2010 in a study
of biomarkers for early disease, PBMC, serum, and SF sam-
ples were obtained during the first year of the disease. All RA
patients had positive test results for rheumatoid factor or
anti–cyclic citrullinated peptide antibodies. PBMCs were
stored in liquid nitrogen, and serum samples were stored at
Isolation and identification of synovial HLA–DR–
presented peptides. A detailed description of the isolation and
identification of in vivo HLA–DR–presented peptides from
patients’ synovial tissue has been published previously (28).
T cell proliferation assay. T cell proliferation assays
were performed as previously described (29). Briefly, patients’
PBMCs were stimulated for 5 days with 2 ?M of each peptide,
after which3H-thymidine was added. All nonredundant HLA–
DR–presented peptides were synthesized by Mimotopes.
Enzyme-linked immunospot (ELISpot) assay. Assays
were performed using an ELISpotpluskit for human IFN?
(Mabtech). ECGF peptides were synthesized and purified by
high-performance liquid chromatography at the Tufts Univer-
sity Core Facility. The peptide sequences were as follows, with
the predicted promiscuous HLA–DR binding peptides shown
in boldface: A52DIRGFVAAVVNGSAQGAQI71; D123KVSLVL-
AUTOIMMUNITY TO ECGF IN LYME DISEASE187
APALAACG137; S220KKLVEGLSALVVDV234; K253TLVGV-
L302RDLVTTLGGALLWL316; and G387TVELVRALPLALV-
LH401. All peptides (1 ?M) were tested in duplicate wells, as
were positive (phytohemagglutinin) and negative (no antigen)
controls. After 5 days, cells were transferred to ELISpot plates
coated with IFN? antibody and incubated overnight. Images of
wells were captured using an ImmunoSpot series 3B analyzer,
and spots were counted using ImmunoSpot software. A posi-
tive T cell response was defined as a stimulation index (SI)
that was ?3 SD above the mean in 18 healthy subjects; this
mean ? 3 SD value characteristically corresponded to ?40
spot-forming units/106PBMCs and an SI of ?8.
Enzyme-linked immunosorbent assay (ELISA) for se-
rum IgG anti-ECGF antibodies. ELISA plates were coated
with 0.5 ?g/ml of carrier-free, recombinant human platelet-
derived EGCF (PD-ECGF; R&D Systems) dissolved in phos-
phate buffered saline (PBS) and incubated overnight at 4°C.
All subsequent steps were performed at room temperature
with plates on a platform shaker set at 200 revolutions per
minute. After washing with PBS–0.05% Tween 20 (PBST), the
plates were incubated with blocking buffer (5% nonfat dry
milk in PBST), followed by patient or control serum samples
(100 ?l; 1:100). Subsequently, horseradish peroxidase–
conjugated goat anti-human IgG was added, followed by TMB
substrate. For interplate standardization, 4 patient and 8
control samples were included on each plate.
Immunoblotting. Human recombinant PD-ECGF was
subjected to polyacrylamide gel electrophoresis and trans-
ferred onto nitrocellulose membranes. Membranes were cut
into strips and incubated for 1 hour in blocking buffer (5%
nonfat dry milk dissolved in 0.1% Tween 20 in 20 mM Tris and
500 mM NaCl; pH 7.5). The strips were incubated with patient
or control sera (1:100), followed by alkaline phosphatase–
conjugated goat anti-human IgG antibody, then visualized with
nitroblue tetrazolium/BCIP substrate. Densitometric analysis
of scanned blots was performed using ImageJ software. To
correct for external variation between experiments, the data
were standardized to a positive control included in each blot.
ELISA for synovial fluid ECGF. ELISA plates were
coated with goat anti-human PD-ECGF (5 ?g/ml) and incu-
bated overnight at 4°C. The next day, plates were incubated
with blocking buffer (5% nonfat dry milk in PBST), followed
by each patient’s SF sample (100 ?l; 1:10). To generate a
standard curve, serial dilutions of PD-ECGF were added to
each plate. The plates were then incubated with mouse anti-
human PD-ECGF antibody, followed by horseradish peroxidase–
conjugated goat anti-mouse IgG and then TMB substrate.
Immunohistochemistry. Synovial tissue samples were
cut and fixed in cold acetone. Endogenous peroxidases were
blocked with hydrogen peroxide in methanol, followed by
1-time power block solution containing 10% normal donkey
serum. The sections were then incubated with anti-rabbit
polyclonal ECGF (3 ?g/ml) at 4°C overnight. For negative con-
trols, nonspecific rabbit IgG was used. The following day, the
sections were incubated with biotinylated anti-rabbit second-
ary antibody, peroxidase–streptavidin, and then diaminobenzi-
dine substrate. The sections were counterstained with Mayer’s
hematoxylin and mounted with glycerol. Microscopic images
were obtained with a Nikon Eclipse ME6000 microscope. Each
slide was read under blinded conditions, and the intensity of
ECGF staining on each tissue region was graded on a semi-
quantitative scale of 0–3, where 0 ? no ECGF-positive cells;
1 ? few (?50) positive cells; 2 ? many (?50–100) positive
cells; and 3 ? most cells (?100) positive.
Statistical analysis. Categorical data were analyzed
using either the chi-square test or Fisher’s exact test, and
quantitative data were analyzed by the Mann-Whitney test. All
analyses were performed using SigmaStat 3.0.
Identification of HLA–DR peptides in synovial
tissue. Based on the hypothesis that HLA–DR molecules
in inflamed synovial tissue, the target of the immune
response in antibiotic-refractory Lyme arthritis, present
disease-related autoantigens, we used a broadly applica-
ble, unbiased approach for the identification of autoan-
tigens in this tissue. The protocol consisted of the
following 3 steps: 1) a proteomics approach using tan-
dem mass spectrometry (MS/MS) for the identification
of HLA–DR–presented peptides in an individual pa-
tient’s synovial tissue, 2) synthesis and testing of all
nonredundant peptides identified for T cell autoreactiv-
ity with PBMCs from the same patient, and 3) deter-
mination of whether any autoreactive peptides and
their source proteins identified in a single patient also
induced T and B cell responses in large numbers of
patients with Lyme disease.
In step 1, we initially analyzed synovial tissue
from 4 patients, 2 with antibiotic-refractory Lyme arthri-
tis, and for comparison, 2 with RA (28). The approach
is shown in Figure 1. Altogether, we identified 1,427
in vivo synovial HLA–DR–presented peptides (220–464
per patient), which were derived from 166 source pro-
teins, including a wide range of intracellular and plasma
proteins. These source proteins were substantially dif-
ferent from those identified from Epstein-Barr virus
cell lines (30), demonstrating the necessity of using
patients’ target tissues or cells for identifying naturally
presented HLA–DR epitopes. Complete lists of pep-
tides, their spectra, and their source proteins for each of
the 4 patients have been published previously (28).
Screening of HLA–DR–presented peptides for
autoantigenicity. In step 2, we first tested peptides
identified from a patient who had a synovectomy for
the treatment of antibiotic-refractory Lyme arthritis
(referred to as LA1 in ref. 28). He had one of the refrac-
tory arthritis–associated HLA–DR alleles (DRB1*
0101). Of the 2,237 MS/MS spectra generated from his
tissue sample, 464 had a consensus match in 2 or more
mass spectrometry search programs (Mascot, OMSSA,
or X!Tandem), of which 104 were nonredundant. Since
we wanted to test as many candidate peptides as possi-
ble, we also manually inspected the 53 peptides identi-
188DROUIN ET AL
fied with only 1 of the 3 search programs, of which 16
could be verified. Altogether, we tested 120 nonre-
dundant peptides for autoreactivity with the patient’s
PBMCs in T cell proliferation assays. Because of limited
cell numbers, peptides were pooled (3 per well) for
Only 2 peptide sets (sets 33 and 40) induced
proliferative responses that were ?2 times background
(Figure 2). When we retested the 6 peptides from these
2 sets, only 1 peptide from set 40 induced a proliferative
response that was ?2 times background, which was
substantially higher than the response observed with any
other peptide. The mass spectrum for this peptide
(L340GRFERMLAAQGVDPG355) is shown in Figure
1E. This peptide was 1 of the 16 peptides identified
with only 1 of the 3 search programs and originated from
the source protein ECGF, also called thymidine phos-
phorylase or gliostatin. ECGF is a chemotactic factor, it
has a proliferative effect on endothelial cells, and it
induces angiogenesis (31). Moreover, it is not known to
be a relevant autoantigen in any other autoimmune
T cell responses to ECGF. In step 3, all available
PBMC and serum samples collected over a 25-year
Figure 1. An overview of the isolation and identification of in vivo HLA–DR–presented peptides from patients’ synovial tissue. A, Antibiotic-
refractory Lyme arthritis usually manifests as one swollen knee. B, In those cases in which therapeutic arthroscopic synovectomies are performed,
20–60 gm of inflamed synovial tissue and subcutaneous fat is removed. C, Immunohistologic staining of the synovial tissue shows marked exogenous
expression of HLA–DR molecules. IHC ? immunohistochemistry. D, HLA–DR complexes are immunoprecipitated (IP) from synovial cell lysates.
E, HLA–DR–presented peptides are eluted and identified by liquid chromatography tandem mass spectrometry (LC-MS/MS). The LC-MS/MS
spectrum of the endothelial cell growth factor (ECGF) peptide ECGF340–355is shown.
Figure 2. Screening of in vivo HLA–DR–presented peptides identi-
fied from the synovial tissue of one patient for T cell autoreactivity
using the patient’s own peripheral blood mononuclear cells. All 120
nonredundant HLA–DR–presented peptides identified were synthe-
sized and tested in sets of 3 (2 ?M of each peptide). T cell proliferation
was measured using a standard3H-thymidine incorporation assay. A
positive result was defined as a proliferative response ?2 times
background (no antigen).
AUTOIMMUNITY TO ECGF IN LYME DISEASE 189
period from patients with EM, the initial skin lesion of
early Lyme disease, and from those with Lyme arthritis,
were tested for T and B cell reactivity with ECGF. For
comparison, samples from healthy control subjects and
from patients with RA were tested. Although HLA–DR
typing was performed only in patients with Lyme arthri-
tis, it is likely that positivity for the HLA–DRB1*0101
and 0401 alleles would be increased among patients
with RA, as in patients with antibiotic-refractory Lyme
Initially, patients’ PBMCs were tested for T cell
autoreactivity using commercially available recombinant
ECGF. However, we found, as had others (32), that
ECGF inhibited the readout of the3H-thymidine assay,
and nonspecifically induced PBMCs to secrete IFN?.
Therefore, using 3 HLA–DR T cell epitope pre-
diction algorithms (33,34), 7 T cell peptide epitopes of
ECGF were identified and synthesized, including the
peptide initially isolated from the patient’s synovial
tissue sample (ECGF340–355). Five of the 7 peptides,
including ECGF340–355, were predicted to be promiscu-
ous HLA–DR binders (predicted to bind ?19 HLA–DR
molecules). Rather than proliferation assays, IFN?
ELISpot assays were used due to their increased sensi-
Healthy control subjects and RA patients had
only minimal responses to a few of the 7 ECGF peptides
tested (Figure 3). Of the patients with EM, an early
manifestation of Lyme disease, 16% had low-level T cell
responses. In contrast, 30% of the patients with antibiotic-
responsive arthritis and 38% of those with antibiotic-
refractory arthritis had robust responses, often to multiple
ECGF peptides. Overall, patients with Lyme arthritis
had substantially greater T cell responses to ECGF pep-
tides than those in the other groups; their cells recog-
nized all 7 peptides tested, and 10 patients had responses
to 2–4 ECGF peptides, suggestive of epitope spreading.
Previously, we showed that patients with antibiotic-
refractory arthritis more often had HLA–DRB1*0101,
0102, 0401, 0402, 0404, or 1501 alleles (15). In our
current study, of 21 patients with refractory or respon-
sive arthritis who had T cell reactivity with ECGF
peptides, 20 (95%) had known refractory arthritis–
associated alleles. Therefore, T cell responses to ECGF
appear to occur primarily in patients with antibiotic-
refractory risk alleles.
Figure 3. Testing of peripheral blood mononuclear cells (PBMCs) from healthy control subjects and patients with rheumatoid arthritis (RA),
erythema migrans (EM), antibiotic-responsive Lyme arthritis, or antibiotic-refractory Lyme arthritis for T cell recognition of endothelial cell growth
factor (ECGF) peptides. PBMCs were incubated with individual ECGF peptides (1 ?M), phytohemagglutinin (positive control), or no peptide
(negative control), and assayed by interferon-? enzyme-linked immunospot assay. The results were quantified using a stimulation index as described
in Patients and Methods, along with the sequences of the peptides. A positive response was defined as a stimulation index of ?3 SD above the mean
in healthy control subjects (the area above the shaded region). The 5 peptides predicted to be promiscuous binders were tested in all patients and
control subjects. Due to limited availability of cells, the 2 peptides predicted to be nonpromiscuous binders (ECGF220–234and ECGF302–316) were
tested in only a subset of patients or control subjects (15 of the 18 healthy control subjects, none of the 12 RA patients, 18 of the 19 patients with
EM, 7 of the 27 patients with antibiotic-responsive arthritis, and 11 of the 37 patients with antibiotic-refractory arthritis).
190DROUIN ET AL
B cell responses to ECGF. ECGF-reactive CD4?
T cells likely contribute to pathogenicity by providing
help to B cells to produce anti-ECGF autoantibodies.
Therefore, we tested patients’ serum samples for IgG
anti-ECGF antibodies using 2 methods, ELISA and
immunoblotting. When ECGF antibody responses were
determined by ELISA, none of the 74 healthy control
subjects had a positive response (defined as ?3 SD
above the mean value in healthy subjects) (Figure 4A).
In comparison, 15% of the patients with EM (P ?
0.001), 8% of the patients with responsive arthritis (P ?
0.04), and 17% of the patients with refractory arthritis
(P ? 0.0001) had positive responses. In addition, pa-
tients with antibiotic-refractory arthritis tended to have
ECGF autoantibodies more frequently than those with
antibiotic-responsive arthritis (17% versus 8%; P ?
0.09). In contrast, none of the 33 patients with RA had
a positive response. When these sera were tested by
immunoblotting, similar results were obtained (Figure
4B), though immunoblotting was not performed in RA
patients, since not enough serum remained. Moreover,
the results obtained by ELISA and those obtained by
immunoblotting were highly concordant in patients
with refractory arthritis (P ? 0.0001), but not in those
with EM or responsive arthritis. These 2 methods may
not assess all of the same epitopes, and therefore,
concordance may occur only with recognition of multiple
When concordance was assessed between T cell
(ELISpot) and B cell (ELISA) assays, 13 of 22 patients
(59%) with responsive arthritis and 23 of 37 patients
(62%) with refractory arthritis had concordant results,
and similar results were obtained when the immuno-
blotting data were compared. Although the overall fre-
quencies of T and B cell immune responses in patients
with EM were similar, concordance was difficult to show
due to the small number of patients with positive T cell
responses (n ? 3). T or B cell reactivity to ECGF did not
correlate with the duration of arthritis or how long the
sample had been frozen prior to testing. Taken together,
these findings indicate that T and B cell responses to
ECGF occurred in patients with early or late manifesta-
tions of Lyme disease, most frequently in those with
Antibody responses to ECGF in patients not
treated with antibiotics. In a study of the natural history
of Lyme disease carried out in the late 1970s, prior
to knowledge of the etiology of the illness, 55 non–
antibiotic-treated patients with EM were followed up
longitudinally for a median of 6 years (6). Of the 55
patients, 21 (38%) had no subsequent manifestations or
only brief joint pain, whereas 34 (62%) subsequently
developed intermittent attacks or persistent arthritis,
lasting from 2 weeks to 4 years. Serial serum samples
were still available for 42 of the 55 patients. Of the 15
patients who did not develop arthritis, 2 (13%) had
ECGF antibody responses 2–3 weeks after the onset of
EM, whereas 7 of the 27 patients (26%) who later had
arthritis had positive ECGF antibody responses, a non-
significant difference. In 6 of the 7 arthritis patients who
Figure 4. IgG anti–endothelial cell growth factor (anti-ECGF) auto-
antibody responses in the sera of healthy control (HC) subjects, and
patients with erythema migrans (EM), antibiotic-responsive Lyme
arthritis, antibiotic-refractory Lyme arthritis, or rheumatoid arthritis
(RA), as determined by enzyme-linked immunosorbent assay (ELISA)
and immunoblotting. A, For the ELISA, plates were coated with
recombinant human ECGF and then incubated with serum samples
(diluted 1:100). B, For immunoblotting, recombinant human ECGF
was subjected to polyacrylamide gel electrophoresis and transferred
to nitrocellulose membranes. Membrane strips were then incubated
with serum samples (diluted 1:100). Band intensity was determined by
densitometry. OD ? optical density.
AUTOIMMUNITY TO ECGF IN LYME DISEASE191
had ECGF responses, reactivity developed weeks to
months after disease onset, prior to joint inflammation.
When patients’ attacks of arthritis were added together,
the duration of active arthritis was significantly longer
in the 7 patients who had ECGF responses than in the
20 patients who did not (median 67 weeks versus 17
weeks, respectively; P ? 0.004). Figure 5 shows the
correlation of disease activity with ECGF antibody levels
in the patient who had the most prolonged arthritis.
Thus, in untreated patients, ECGF antibody responses
usually developed early in the illness, and in those who
subsequently developed arthritis, this response was as-
sociated with a significantly longer duration of joint
ECGF in joint fluid and synovial tissue. For
ECGF to have pathogenic relevance as an autoantigen
in antibiotic-refractory Lyme arthritis, one would pre-
dict that this protein would be present in high concen-
trations in patients’ inflamed SF and tissue. Although SF
samples were available for patients with antibiotic-
responsive arthritis, synovial tissue was not, since thera-
peutic synovectomies are never necessary in this patient
group. As determined by sandwich ELISA, patients
with antibiotic-refractory arthritis often had very high
concentrations of ECGF in SF (mean 448 ng/ml)
(Figure 6A), which were significantly greater than those
in patients with antibiotic-responsive arthritis (mean 154
ng/ml) (P ? 0.0001). RA patients also often had high
levels of ECGF (mean 313 ng/ml), which is consistent
Figure 6. Detection of endothelial cell growth factor (ECGF) protein
in joint fluid and synovial tissue. A, ECGF concentrations in joint fluid
from patients with antibiotic-responsive Lyme arthritis, antibiotic-
refractory Lyme arthritis, or rheumatoid arthritis (RA) were measured
by sandwich enzyme-linked immunosorbent assay. Symbols represent
individual patients; horizontal lines represent the mean. B, Represen-
tative serial synovial tissue sections from a patient with antibiotic-
refractory Lyme arthritis and a patient with RA stained with anti-
ECGF or isotype control antibodies. Insets show higher-magnification
views of the boxed areas. Circles indicate the sublining area around
blood vessels; arrows indicate large cells, most likely synoviocytes.
Figure 5. Correlation of the clinical course with IgG anti–endothelial
cell growth factor (anti-ECGF) antibody responses as determined by
enzyme-linked immunosorbent assay (ELISA) in a historical patient
not treated with antibiotics and followed up for 6 years throughout his
entire disease course. He initially had erythema migrans, followed
several months later by a brief episode of arthritis, and then prolonged
arthritis in a hip for 4 years. The antibody response to ECGF
correlated with the 2 episodes of arthritis, and was highest during the
period of prolonged arthritis. The shaded region represents the range
of values found in healthy control subjects, as shown in Figure 4 (with
the top of the range being 3 SD above the mean in the control subjects,
and levels above this considered positive). OD ? optical density.
192 DROUIN ET AL
with the findings of other investigators (32). Those
investigators also showed that patients with osteoarthri-
tis, a minimally inflammatory form of arthritis, had
much lower levels (mean 8.7 ng/ml).
Synovial tissue specimens from 16 patients with
antibiotic-refractory arthritis and 5 patients with RA
were examined for the presence of ECGF. Of the 16
patients with antibiotic-refractory arthritis, 10 (63%)
had moderate-to-intense staining for ECGF in the lining
and sublining of the synovial tissue, 4 (25%) had mild
staining, and 2 (12%) had no staining in these areas. In
comparison, the 5 RA patients had staining for ECGF
primarily in the lining area, with little seen in the
sublining region. Representative examples from a pa-
tient with antibiotic-refractory arthritis and a patient
with RA are shown in Figure 6B. In the synovial tissue of
the patient with antibiotic-refractory arthritis, ECGF
staining was clearly evident in the sublining area around
blood vessels (green circles) and in large cells, most
likely synoviocytes (red arrows). In contrast, in the RA
patients, staining was not seen in the sublining region.
Thus, the majority of patients with antibiotic-refractory
arthritis had large amounts of ECGF in SF and intense
staining in synovial tissue, where the protein could act as
In this study, we used discovery-based proteomics
to identify HLA–DR–presented self peptides in a pa-
tient’s synovium and determined the immunoreactivity
of the peptides by testing them with his PBMCs. One
peptide derived from ECGF was shown to be auto-
reactive. We then found that ?10–30% of patients
with early or late manifestations of Lyme disease had
T or B cell responses to ECGF, and they were more
common in patients with antibiotic-refractory arthritis.
As seen in non–antibiotic-treated historical patients,
ECGF antibody responses often developed early in the
illness, prior to the onset of arthritis, analogous to the
situation in RA or lupus (3,4). Moreover, ECGF anti-
body responses in these patients were associated with
more persistent arthritis. Previously, naturally presented
HLA–DR peptides had been identified from other
tissues and fluids by MS/MS (35–39), but until our recent
study (28), not from synovial tissue. Additionally, no
previous studies had systematically tested the autoreac-
tivity of each peptide identified using PBMCs from the
same patient, the key step in autoantigen identification
in our present study.
ECGF itself was found in high levels in SF and
synovial tissue from patients with antibiotic-refractory
Lyme arthritis and those with RA. However, the protein
appeared to be immunoreactive primarily in patients
with Lyme disease. The reasons why Bb infection may
lead to ECGF reactivity are not yet known. T cell
epitope mimicry between a spirochete and host protein
is often the first hypothesis considered. In a search of the
Bb proteome, the Bb peptide with the greatest sequence
homology with ECGF340–355was derived from BB_0580
(F78ERMLA83), a Bb protein that is not known to be
immunoreactive (40). This Bb peptide has sequence
homology with 6 of 9 ECGF340–355core residues, but in
our experience (18), this is insufficient to induce cross-
reactivity. Moreover, contrary to what one would expect
with a molecular mimicry mechanism, no single ECGF
epitope was recognized in all, or even the majority of,
Alternatively, ECGF antibody responses could
reflect a cross-reactive antibody response to Bb. How-
ever, the fact that patients had both T and B cell
responses to ECGF, including responses to multiple
T cell epitopes, is evidence against this possibility.
Another option is that excessive joint inflammation may
lead to bystander activation of ECGF-specific T cells.
However, the lack of robust ECGF reactivity in RA
patients, despite high levels of the protein in joints, is
evidence against this type of nonspecific mechanism.
One additional possibility is that Bb, which is known
to bind several host proteins (41,42), may also bind
ECGF, allowing Bb to act simultaneously as a conduit
for enhanced presentation by antigen-presenting cells
(APCs) and as an adjuvant.
We postulate a 3-step process in the pathogenesis
of antibiotic-refractory Lyme arthritis. In the first step,
autoimmune responses to ECGF develop systemically in
?10–20% of individuals with Lyme disease, often early
in the illness. Both spirochete and host genetic factors
are probably important in the generation of this auto-
immune response. For example, Bb RST1 strains, which
account for ?30–50% of the infections in the north-
eastern US, are more inflammatory than other Bb
strains (43), and more frequently cause antibiotic-
refractory arthritis (44). Important host factors likely
include specific HLA genotypes, such as DRB1*0101 or
0401 (15), which are also found more often in patients
who develop antibiotic-refractory arthritis.
In the second step, we propose that this rather
common systemic autoimmune response becomes
pathogenic in only the small percentage of patients who
have marked ECGF antigen accumulation, excessive
inflammation, and immune dysregulation in joints. Pa-
AUTOIMMUNITY TO ECGF IN LYME DISEASE193
tients with refractory arthritis had elevated concentra-
tions of ECGF in their SF, where it could be taken up
by local APCs and presented at high concentrations
to T cells, leading to their activation. Moreover, intense
staining for ECGF was detected in the synovia of
patients with refractory arthritis, where it could initiate
immune complex deposition and tissue damage through
excessive complement activation. However, more than
one autoantigen or certain spirochetal antigens may be
important in antibiotic-refractory arthritis, since a con-
siderable number of patients with refractory arthritis
do not have anti-ECGF antibodies. Moreover, in un-
treated patients, infection and autoimmunity may occur
together, though the arthritis seems to be more persis-
tent when the autoimmune component is present. It is
only with antibiotic treatment that one may observe the
autoimmune component independently.
Even though more than one triggering antigen
may be involved, patients with antibiotic-refractory ar-
thritis, particularly those with a TLR1 polymorphism
(1805GG) (45), have exceptionally high levels of pro-
inflammatory mediators, such as tumor necrosis factor
?, interleukin-1?, and IFN?, in SF and synovial tissue
compared to patients with responsive arthritis (14).
Although these inflammatory mediators are initially
important for eradication of the spirochete, the inability
to down-regulate their expression in these individuals
likely contributes to immune dysregulation (46,47). Fi-
nally, in patients with antibiotic-refractory arthritis,
CD4? T effector cells in SF resist suppression by CD4?
Treg cells, and lower numbers of CD4? Treg cells cor-
relate with longer durations of arthritis (48).
In the third step, synovitis in most patients re-
solves within months to several years after antibiotic
therapy, assisted by disease-modifying antirheumatic
drugs such as methotrexate, which are thought to inhibit
T cell activation (49). In these patients, we postulate
that the innate immune “danger” signals provided by
live spirochetes are no longer present, and without these
signals, the adaptive immune response to autoantigens
eventually regains homeostasis. Similarly, in patients
requiring synovectomies, the arthritis does not usually
recur because innate immune signals associated with
active infection are missing.
In summary, we have shown definitively that T
and B cell responses to ECGF occur in a subset of
patients with Lyme disease, thereby identifying the first
autoantigen that is a target of autoimmune T and
B cell responses in this illness. Moreover, the approach
used here for the identification of novel autoreactive
HLA—DR–presented peptides in the synovia of pa-
tients with antibiotic-refractory Lyme arthritis should
be valuable for the determination of immune targets
in other forms of chronic inflammatory arthritis, includ-
The authors thank Drs. Nitin Damle and Vijay Sikand
for the collection of samples from patients with EM, Drs.
Elena Massarotti and Robert Kalish for help in obtaining
samples from patients with Lyme arthritis, and Dr. Lisa
Glickstein for helpful discussion about the manuscript.
All authors were involved in drafting the article or revising it
critically for important intellectual content and all authors approved
the final version to be published. Dr. Drouin 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 concept and design. Drouin, Costello, Steere.
Acquisition of data. Drouin, Seward, Strle, McHugh, Katchar,
London ˜o, Yao.
Analysis and interpretation of data. Drouin, Costello, Steere.
1. Sonderstrup G, McDevitt HO. DR, DQ, and you: MHC alleles
and autoimmunity. J Clin Invest 2001;107:795–6.
2. Cho JH, Gregersen PK. Genomics and the multifactorial nature of
human autoimmune disease. N Engl J Med 2011;365:1612–23.
3. McInnes IB, Schett G. The pathogenesis of rheumatoid arthritis.
N Engl J Med 2011;365:2205–19.
4. Arbuckle MR, McClain MT, Rubertone MV, Scofield RH, Dennis
GJ, James JA, et al. Development of autoantibodies before the
clinical onset of systemic lupus erythematosus. N Engl J Med
5. Steere AC. Lyme disease. N Engl J Med 2001;345:115–25.
6. Steere AC, Schoen RT, Taylor E. The clinical evolution of Lyme
arthritis. Ann Intern Med 1987;107:725–31.
7. Steere AC, Levin RE, Molloy PJ, Kalish RA, Abraham JH III,
Liu NY, et al. Treatment of Lyme arthritis. Arthritis Rheum
8. Steere AC, Angelis SM. Therapy for Lyme arthritis: strategies for
the treatment of antibiotic-refractory arthritis [review]. Arthritis
9. Steere AC, Glickstein L. Elucidation of Lyme arthritis. Nat Rev
10. Bockenstedt LK, Gonzalez DG, Haberman AM, Belperron AA.
Spirochete antigens persist near cartilage after murine Lyme
borreliosis therapy. J Clin Invest 2012;122:2652–60.
11. Li X, McHugh GA, Damle N, Sikand VK, Glickstein L, Steere AC.
Burden and viability of Borrelia burgdorferi in skin and joints of
patients with erythema migrans or Lyme arthritis. Arthritis Rheum
12. Kannian P, Drouin EE, Glickstein L, Kwok WW, Nepom GT,
Steere AC. Decline in the frequencies of Borrelia burgdorferi
OspA161-175-specific T cells after antibiotic therapy in HLA-
DRB1*0401-positive patients with antibiotic-responsive or anti-
biotic-refractory Lyme arthritis. J Immunol 2007;179:6336–42.
13. Kannian P, McHugh G, Johnson BJ, Bacon RM, Glickstein LJ,
Steere AC. Antibody responses to Borrelia burgdorferi in patients
with antibiotic-refractory, antibiotic-responsive, or non–antibiotic-
treated Lyme arthritis. Arthritis Rheum 2007;56:4216–25.
194 DROUIN ET AL
14. Shin JJ, Glickstein L, Steere AC. High levels of inflammatory
chemokines and cytokines in joint fluid and synovial tissue through-
out the course of antibiotic-refractory Lyme arthritis. Arthritis
15. Steere AC, Klitz W, Drouin EE, Falk BA, Kwok WW, Nepom GT,
et al. Antibiotic-refractory Lyme arthritis is associated with
HLA-DR molecules that bind a Borrelia burgdorferi peptide.
J Exp Med 2006;203:961–71.
16. Steere AC, Duray PH, Butcher EC. Spirochetal antigens and
lymphoid cell surface markers in Lyme synovitis: comparison
with rheumatoid synovium and tonsillar lymphoid tissue. Arthritis
17. Gross DM, Forsthuber T, Tary-Lehmann M, Etling C, Ito K, Nagy
ZA, et al. Identification of LFA-1 as a candidate autoantigen in
treatment-resistant Lyme arthritis. Science 1998;281:703–6.
18. Drouin EE, Glickstein L, Kwok WW, Nepom GT, Steere AC.
Human homologues of a Borrelia T cell epitope associated
with antibiotic-refractory Lyme arthritis. Mol Immunol 2008;45:
19. Steere AC, Drouin EE, Glickstein LJ. Relationship between
immunity to Borrelia burgdorferi outer-surface protein A (OspA)
and Lyme arthritis. Clin Infect Dis 2011;52 S259–65.
20. Steere AC, Falk B, Drouin EE, Baxter-Lowe LA, Hammer J,
Nepom GT. Binding of outer surface protein A and human
lymphocyte function–associated antigen 1 peptides to HLA–DR
molecules associated with antibiotic treatment–resistant Lyme
arthritis. Arthritis Rheum 2003;48:534–40.
21. Ghosh S, Seward R, Costello CE, Stollar BD, Huber BT. Auto-
antibodies from synovial lesions in chronic, antibiotic treatment-
resistant Lyme arthritis bind cytokeratin-10. J Immunol 2006;177:
22. Martin R, Ortlauf J, Sticht-Groh V, Bogdahn U, Goldmann SF,
Mertens HG. Borrelia burgdorferi–specific and autoreactive T-cell
lines from cerebrospinal fluid in Lyme radiculomyelitis. Ann
23. Kuenzle S, von Budingen HC, Meier M, Harrer MD, Urich E,
Becher B, et al. Pathogen specificity and autoimmunity are distinct
features of antigen-driven immune responses in neuroborreliosis.
Infect Immun 2007;75:3842–7.
24. Lunemann JD, Gelderblom H, Sospedra M, Quandt JA, Pinilla C,
Marques A, et al. Cerebrospinal fluid-infiltrating CD4?T cells
recognize Borrelia burgdorferi lysine-enriched protein domains
and central nervous system autoantigens in early Lyme encepha-
litis. Infect Immun 2007;75:243–51.
25. Chandra A, Wormser GP, Klempner MS, Trevino RP, Crow MK,
Latov N, et al. Anti-neural antibody reactivity in patients with a
history of Lyme borreliosis and persistent symptoms. Brain Behav
26. Wharton M, Chorba TL, Vogt RL, Morse DL, Buehler JW. Case
definitions for public health surveillance. MMWR Recomm Rep
27. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham
CO III, et al. 2010 rheumatoid arthritis classification criteria: an
American College of Rheumatology/European League Against
Rheumatism collaborative initiative. Arthritis Rheum 2010;62:
28. Seward RJ, Drouin EE, Steere AC, Costello CE. Peptides pre-
sented by HLA-DR molecules in synovia of patients with rheu-
matoid arthritis or antibiotic-refractory Lyme arthritis. Mol Cell
29. Drouin EE, Glickstein L, Kwok WW, Nepom GT, Steere AC.
Searching for borrelial T cell epitopes associated with antibiotic-
refractory Lyme arthritis. Mol Immunol 2008;45:2323–32.
30. Chicz RM, Urban RG, Gorga JC, Vignali DA, Lane WS,
Strominger JL. Specificity and promiscuity among naturally pro-
cessed peptides bound to HLA-DR alleles. J Exp Med 1993;178:
31. Ishikawa F, Miyazono K, Hellman U, Drexler H, Wernstedt C,
Hagiwara K, et al. Identification of angiogenic activity and the
cloning and expression of platelet-derived endothelial cell growth
factor. Nature 1989;338:557–62.
32. Takeuchi M, Otsuka T, Matsui N, Asai K, Hirano T, Moriyama A,
et al. Aberrant production of gliostatin/platelet-derived endothe-
lial cell growth factor in rheumatoid synovium. Arthritis Rheum
33. Sturniolo T, Bono E, Ding J, Raddrizzani L, Tuereci O, Sahin U,
et al. Generation of tissue-specific and promiscuous HLA ligand
databases using DNA microarrays and virtual HLA class II
matrices. Nat Biotechnol 1999;17:555–61.
34. Wang P, Sidney J, Dow C, Mothe B, Sette A, Peters B. A
systematic assessment of MHC class II peptide binding predictions
and evaluation of a consensus approach. PLoS Comput Biol
35. Gordon RD, Young JA, Rayner S, Luke RW, Crowther ML,
Wordsworth P, et al. Purification and characterization of endog-
enous peptides extracted from HLA-DR isolated from the spleen
of a patient with rheumatoid arthritis. Eur J Immunol 1995;25:
36. Muixi L, Carrascal M, Alvarez I, Daura X, Marti M, Armengol
MP, et al. Thyroglobulin peptides associate in vivo to HLA-DR in
autoimmune thyroid glands. J Immunol 2008;181:795–807.
37. Oshitani N, Hato F, Kitagawa S, Maeda K, Higuchi K, Matsumoto
T, et al. Analysis of intestinal HLA-DR bound peptides and
dysregulated immune responses to enteric flora in the pathogen-
esis of inflammatory bowel disease. Int J Mol Med 2003;11:
38. Wahlstrom J, Dengjel J, Persson B, Duyar H, Rammensee HG,
Stevanovic S, et al. Identification of HLA-DR-bound peptides
presented by human bronchoalveolar lavage cells in sarcoidosis.
J Clin Invest 2007;117:3576–82.
39. Fissolo N, Haag S, de Graaf KL, Drews O, Stevanovic S, Ram-
mensee HG, et al. Naturally presented peptides on major histo-
compatibility complex I and II molecules eluted from central
nervous system of multiple sclerosis patients. Mol Cell Proteomics
40. Barbour AG, Jasinskas A, Kayala MA, Davies DH, Steere AC,
Baldi P, et al. A genome-wide proteome array reveals a limited set
of immunogens in natural infections of humans and white-footed
mice with Borrelia burgdorferi. Infect Immun 2008;76:3374–89.
41. Coleman JL, Gebbia JA, Piesman J, Degen JL, Bugge TH, Benach
JL. Plasminogen is required for efficient dissemination of B.
burgdorferi in ticks and for enhancement of spirochetemia in mice.
42. Hallstrom T, Haupt K, Kraiczy P, Hortschansky P, Wallich R,
Skerka C, et al. Complement regulator-acquiring surface protein 1
of Borrelia burgdorferi binds to human bone morphogenic protein
2, several extracellular matrix proteins, and plasminogen. J Infect
43. Strle K, Jones KL, Drouin EE, Li X, Steere AC. Borrelia
burgdorferi RST1 (OspC type A) genotype is associated with
greater inflammation and more severe Lyme disease. Am J Pathol
44. Jones KL, McHugh GA, Glickstein LJ, Steere AC. Analysis of
Borrelia burgdorferi genotypes in patients with Lyme arthritis:
high frequency of ribosomal RNA intergenic spacer type 1 strains in
antibiotic-refractory arthritis. Arthritis Rheum 2009;60:2174–82.
45. Strle K, Shin JJ, Glickstein LJ, Steere AC. Association of a
Toll-like receptor 1 polymorphism with heightened Th1 inflamma-
tory responses and antibiotic-refractory Lyme arthritis. Arthritis
46. Van Amelsfort JM, van Roon JA, Noordegraaf M, Jacobs KM,
Bijlsma JW, Lafeber FP, et al. Proinflammatory mediator–induced
reversal of CD4?,CD25? regulatory T cell–mediated suppression
in rheumatoid arthritis. Arthritis Rheum 2007;56:732–42.
AUTOIMMUNITY TO ECGF IN LYME DISEASE 195
47. Herrath J, Muller M, Amoudruz P, Janson P, Michaelsson J,
Larsson PT, et al. The inflammatory milieu in the rheumatic
joint reduces regulatory T-cell function. Eur J Immunol 2011;41:
48. Shen S, Shin JJ, Strle K, McHugh G, Li X, Glickstein LJ, et al.
Treg cell numbers and function in patients with antibiotic-
refractory or antibiotic-responsive Lyme arthritis. Arthritis Rheum
49. Johnston A, Gudjonsson JE, Sigmundsdottir H, Ludviksson BR,
Valdimarsson H. The anti-inflammatory action of methotrexate is
not mediated by lymphocyte apoptosis, but by the suppression of
activation and adhesion molecules. Clin Immunol 2005;114:154–63.
Clinical Images: Butterfly vertebra
The patient, a 24-year-old woman, presented with a 1-year history of mechanical low back pain without radiation to the legs. The
patient’s medical history and results of the physical examination were unremarkable. A lateral radiograph of the lumbar spine
showed reversal of the physiologic lordosis and reduction in the anterior height of the L4 vertebral body with a cuneiform wedging
appearance (arrow in A) resembling a vertebral collapse. Two L4 hemivertebrae with a complete central cleft (arrows in B) and an
abnormal, complementary L3 vertebral shape were documented by computed tomography (B) and magnetic resonance imaging (C).
These radiographic findings confirmed the diagnosis of butterfly vertebra, a rare congenital malformation that results when the 2
chondrification centers in an embryonic vertebral body do not fuse. This anomaly is also known as sagittal cleft vertebra, anterior
rachischisis, somatoschisis, or anterior spina bifida. It occurs most commonly in the lumbar spine, and it may be associated with
complex congenital syndromes, such as Pfeiffer syndrome, Jarcho-Levin syndrome, Crouzon syndrome, and Alagille syndrome.
Butterfly vertebra is generally asymptomatic and incidentally detected. Hypoplasia of one or both halves of the vertebral body can
lead to lateral displacement or anterior wedging, resulting in kyphosis or scoliosis, and can be a cause of chronic back pain. Although
congenital vertebral anomalies are relatively common (global prevalence 0.5–1/1,000 live births), very few cases of symmetric fusion
defects resulting in butterfly vertebra have been reported. Awareness of this congenital malformation is important, since it may
easily be confused with a vertebral fracture.
University of Alberta
Edmonton, Alberta, Canada
Juan Francisco Asenjo, MD, FRCPC
Ine ´s Colmegna, MD
McGill University Health Centre
Montreal, Quebec, Canada
196 DROUIN ET AL