Identification of Novel Biomarkers for Behcet Disease Diagnosis Using Human Proteome Microarray Approach*

Behcet disease (BD) is a chronic systemic vasculitis and considered as an autoimmune disease. Although rare, BD can be fatal due to ruptured vascular aneurysms or severe neurological complications. To date, no known biomarker has been reported for this disease, making it difficult to diagnosis in the clinics. To undertake this challenge, we employed the HuProt arrays, each comprised of ∼20,000 unique human proteins, to identify BD-specific autoantibodies using a Two-Phase strategy established previously. In Phase I, we profiled the autoimmunity on the HuProt arrays with 75 serum samples collected from 40 BD patients, 15 diagnosed autoimmune patients who suffer from Takayasu arteritis (TA; n = 5)), ANCA associated vasculitis (AAV; n = 5), and Sjogren's syndrome (SS; n = 5), and 20 healthy subjects, and identified 20 candidate autoantigens that were significantly associated with BD. To validate these candidates, in Phase II we constructed a focused array with these 20 candidate BD-associated antigens, and use it to profile a much larger cohort, comprised of serum samples collected from 130 BD patients, 103 autoimmune patients (i.e. 40TA, 40 AAV and 23 SS), and 110 healthy controls. This allowed us to validate CTDP1 (RNA polymerase II subunit A C-terminal domain phosphatase)as a BD-specific autoantigen. The association of anti-CTDP1 with BD patients was further validated using the traditional Western blotting analysis. In conclusion, anti-CTDP1 antibody serves a novel autoantibody for Behcet disease and is expected to help more accurate clinical diagnosis.

antibody (ANA), antiphospholipid antibody (APLA), and antineutrophil cytoplasmic antibodies (ANCA), which are commonly found in many autoimmune diseases. Though the presence of anti-endothelial cell autoantibody (AECA) can partially account for the vascular damage of BD (8 -10), AECA is observed at a low penetration rate of 13.1%, 26%, and 47.5% in Turkish, Spanish, and Chinese BD patients respectively, and its presence in other systemic vasculitis, such as systemic lupus erythematosus (SLE) and systemic sclerosis (SSc), is rather high, making accurate and timely diagnosis very challenging (11)(12)(13). Until now, diagnosis of BD is mainly dependent on clinical criteria. Although there is not a lack of effort to discover ideal BD biomarkers, including a-tropomyosin, selenium binding protein, haptoglobin, amyloidA, cofilin-1 mitochondrial carrier homolog 1and kinectin, via proteomic or express cloning approaches, the impact has been limited (14 -19).
During the past decade, functional protein microarrays, in particular the HuProt arrays, comprised of ϳ20,000 individually purified human proteins, have become a powerful tool for identifying novel biomarkers in many autoimmune diseases, such as rheumatoid arthritis (RA), primary biliary cirrhosis (PBC), SSc, and type 1 diabetes, to name a few (20 -23), because it allows simultaneous profiling autoimmunity against nearly the entire human proteome with minimum consumption of the clinical samples (24). To take advantage of this unbiased proteomics tool, we applied a previously reported twostage strategy to identify BD-specific biomarkers. In Phase I, the HuProt arrays were employed to identify candidate biomarkers using a relatively small cohort, to identify candidate autoantigens. In Phase II, a focused array, comprised of these candidate autoantigens, was employed to validate these autoantigens via screening against a much larger cohort (Fig. 1). On the basis of statistic analysis, CTDP1 was confirmed as a novel BD autoantigen and was further validated by Western blotting analysis.

EXPERIMENTAL PROCEDURES
Serum Samples-All serum samples involved in this study were collected at Peking Union Medical College Hospital during a period from February 2012 to January 2014. This effort involved collecting serum samples from170 BD patients (37.1 Ϯ 11.6 years of age; 79 females), the clinical symptoms information of the BD patient is shown in Table I (25) and were excluded from other autoimmune diseases, such as systemic lupus erythematosus, inflammatory bowel disease, RA, diabetes, and ankylosing spondylitis (AS). Disease controls were diagnosed according to the corresponding general criteria (25)(26)(27). Serum samples were obtained by separation from peripheral blood and stored at Ϫ80°C until use. This study was approved by the Ethics Committee of Peking Union Medical College Hospital.

Construction and Quality Control Test of HuProt Arrays-The
Hu-Prot arrays, comprised ofϳ20,000 unique full-length proteins were constructed in Dr. Zhu's laboratory at Johns Hopkins University School of Medicine. Briefly, human ORFs cloned into a yeast expression vector (pEGH-A) were induced to produce N-terminally tagged GST fusions in Saccharomyces cerevisiae under control of the galactose-inducible GAL1 promoter. Using a high-throughput protein purification protocol, these GST fusions were purified in a 96-well format. Together with negative and positive control probes, all purified human proteins were printed in duplicate onto a single glass slide. The quality of the HuProt arrays was monitored by anti-GST probing, followed by the Cy5-labeled secondary antibodies (Fig. 2). The Hu-Prot arrays were stored at-80°C until use.
Serum Profiling with HuProt Arrays-First, a small cohort of serum samples collected from 40 BD patients,15 autoimmune disease patients (i.e. 5 TA, 5 AAV, and 5 SS), and 20 healthy controls was probed individually to 75 HuProt arrays, the characteristics of study participants involved in this study is shown in Table II, as the first phage autoantibodies profiling assay (Fig. 1). HuProtarrays were taken out

New BD Biomarkers by HuProt Array Approach
from Ϫ80°C and warmed up at room temperature for half an hour and then incubated in a blocking buffer (3% BSA in PBS buffer with 0.1% Tween 20) at 37°C for 1 h. Then a serum sample diluted1:1000 fold into 150 l blocking buffer was added and incubated under coverslip (LifterSlip, Erie Scientific Company, Portsmouth, NH) at 37°C for 1h. After 3 ϫ 10 min washes with PBST, the microarray was incubated with 150 l of 1:1000 diluted Alexa 647 conjugated goat anti-human IgG (the Jackson Laboratory, Bar Harbor, ME) at 37°C for 1h in dark. Finally, after 3 ϫ 10 min PBST washes the microarray was rinsed with double-distilled H 2 O and dried. The microarray was scanned with the GenePix 4000B Microarray Scanner (Molecular Devices, Sunnyvale, CA) and analyzed using GenePix Pro 6.0 software (Molecular Devices).
Construction of BD Focused Microarray and Serum Assay-Together with negative and positive controls, 20 candidate autoantigens potentially associated with BD as determined using the Hu-Prot array data were purified and printed in duplicate in 12 identical sub-arrays on a single OPEpoxy SlideTM to construct the BD focused microarrays. A 12-hole rubber gasket divided each mi-croarray to 12 individual chambers for 12 individual serum assays at same time. These low-density focused BD arrays were stored at Ϫ80°C until use.
To validate these candidate autoantigens in Phase II, a larger cohort of serum samples collected from 130 BD patients, 103 autoimmune disease controls (DC) (i.e. 40TA, 40 AAV and 23 SS), and110 healthy controls were subjected to the focused BD array profiling. The protocol for the serum assay on BD focused microarray was basically identical to that of the HuProt array, except that 50 l of 1/1000diluted human serum and 50L1/1000-diluted goat anti-human IgG were sequentially incubated in each chamber. After carefully removing the rubber gaskets, microarrays were washed and scanned.
Microarray Data Analysis-For HuProt array assays, the median foreground and background intensity for each spot on the arrays were acquired with GenePix Pro 6.0 software. The ratio of foreground to background signals for each spot was considered as the spot's signal value, and then the mean signal value of each duplicate pair was taken as the protein's signal value. We set the signal value cutoff at 2 to identify the positives. Differential proteins were identified using Gene Pattern platform (28). t test was chosen to assess the differential significance for each protein between different groups (BD) based on signal value. BD-specific candidate autoantigens were selected using p value Ͻ0.05 and fold change Ͼ1.3. To ensure appropriate sensitivity, only the antigens that showed a Ͼ20% positive rate (proportion of samples that are identified as positives in the BD group) in all the BD patients were considered as candidate autoantigens that were potentially associated with BD.
For the focused microarrays fabricated with those BD candidate autoantigens, the signal value for each protein was defined by ratio (dividing foreground intensity by the background intensity). Through comparing BD and control groups using t test, the proteins with p  valueϽ0.05 were considered as statistically significant BD-associated autoantigens.
Western Blotting Analysis-CTDP1 were expressed and purified from yeast as GST fusions. After electrophoresis on a 12% SDS-PAGE gel, CTDP1 protein was transferred to PVDF membrane, blocked with 5% nonfat milk, and incubated with 1:200 sera, followed by horseradish peroxidase conjugated anti-human IgG. The immuno reactive bands were visualized by ImageQuant (GE Healthcare).

Identification of Candidate Autoantigens
Associated with BD Using HuProt Arrays-As described previously, in Phase I of the Two Phase strategy, we employed the HuProt arrays to profile a relatively small cohort. We assembled a cohort, comprised of serum samples collected from 40 BD patients, 15 autoimmune disease controls and 20 healthy controls (see Materials and Methods for more details). Each of the serum samples was individually incubated on the HuProt arrays, followed by incubation with a fluorescently labeled secondary anti-human IgG antibody to identify immune-reactive autoantigens. Using a relatively stringent cutoff value, human protein positively reacted to each serum sample were identified. However, number of identified positive hits identified by each serum sample varied dramatically, ranging from ϳ100 to more than 1000 proteins, regardless of the origin of the samples (Fig. 3A). Comparison the number of hit number by each serum sample did not identify any significant difference between the BD patients and healthy control, or between autoimmune disease controls and healthy controls (Fig. 3B).
To identify potential biomarkers for BD diagnosis, data obtained on the HuProt arrays were first processed using Significant Analysis of Microarray (SAM) algorithm. BD-specific candidate autoantigens were identified between 3 groups, BD/Control, BD/Nor and BD/DC based on the p value Ͻ0.05 and fold change Ͼ1.3. We also set an arbitrary penetration rate at Ͼ20% among the BD patient samples, in order to improve the stringency. A total of 20 candidate BD-associated autoantigens were therefore identified (Table III). Using a one-dimensional clustering analysis, the autoimmunity of these 20 proteins can be visualized (Fig. 4A). For example, PRAMEF12 showed a sensitivity and specificity of 32.5% (i.e. 13positives identified among 40 BD samples) and 91.4% (i.e.

New BD Biomarkers by HuProt Array Approach
32negatives among 20 healthy controls and 15 disease controls). As illustrated in Fig. 4B, two representative proteins, namely CTDP1 and PRAMEF12, were specifically recognized by a BD serum sample, but are less likely to be recognized by serum samples either collected from the disease or healthy controls. CTDP1 is a phosphatase that dephosphorylates the C-terminal domain of RNA polymerase II, but little is known about the function of PRAMEF12. None of these proteins are known to be associated with any autoimmune diseases. Furthermore, box plot analysis also demonstrated a significant segregation between BD and control groups by three other proteins, namely FAXDC2, CLPTM1, and NBPF1 (Fig. 4C). Therefore, it was likely that some of these 20 potential biomarkers should be validated in Phase II.
Validation of BD-associated Autoantigens with BD Focused Microarrays-To validate the results obtained from HuProt arrays analysis in Phase II, we cherry-picked 20 yeast strains that each carries one of the 20 human ORFs on an expression vector, pEGH-A, from our human ORF collection. To purify these proteins, the 20 yeast strains were grown in 16 ml of culture, induced by galactose, and purified as N-terminal GST fusion proteins using a standard protocol described previously (29). After examination of the quantity and quality of the purified proteins using Coomassie stain and immunoblot analysis (data not shown), these purified proteins were spotted robotically onto glass sides to form the BD focused arrays. To save the cost, each glass slide contained 12 identical subarrays in a 2 ϫ 6 format. Next, a much larger cohort, comprised of serum samples collected from 130 BD patients, 103 patients diagnosed of other types of autoimmune diseases, and 110 healthy subjects, was assembled and each serum sample was individually incubated on a subarray of the BD focused microarrays. The same assay conditions were applied to these focused arrays and the binding signals acquired, as described above. Using the same data analysis procedure, we validated anti-CTDP1 antibodies as the most significant biomarker that distinguishes BD patients from other autoimmune patients, as well as from healthy subjects (Fig. 5A). In this large cohort of 343 serum samples, anti-CTDP1 antibodies showed 23.86% sensitivity among the BD patients, which is significantly higher than that among disease control group (i.e. 4.9%; X2 ϭ 15.87; p ϭ 6.79 ϫ 10 -5), and than that among the healthy controls (i.e. 6 the results obtained from the BD focused arrays agreed very well with that obtained from the HuProt arrays, as illustrated in boxplot analysis (Fig. 5B).
Detect Autoantibodies to CTDP1 by Western Blotting-To test the potential of transforming anti-CTDP1 antibodies as a more clinically friend biomarker, we purified anti-CTDP1 antibodies proteins and tested it using the traditional Western blotting analysis using serum samples collected from two anti-CTDP1 positive BD patients, as well as samples collected from patients diagnosed for other autoimmune disease, including TA and AAV. Samples from healthy two subjects were also tested as negative controls. As expected, anti-CTDP1 antibodies could only be detected with the two BD serum samples, but not by any other samples (Fig. 5C). Therefore, we believe that anti-CTDP1 antibodies test can be potentially converted to ELISA-based assays in the future. DISCUSSION BD is a rare immune-mediated small-vessel systemic vasculitis and is considered as an autoimmune disease in the clinics. Although a handful of recent studies showed that antiendothelial cell antibody (AECA) was found in serum samples of BD patients, AECA has a very low specificity in the clinics and the association of AECA with BD pathogenesis has not been firmly established. Therefore, it is in urgent need to employ an unbiased, high-throughput method to identify novel autoantibodies in BD patients. Protein microarrays, especially human proteome microarrays, are emerging as a highly effective approach for profiling new autoantibodies in numerous autoimmune diseases, as well as cancers (21,30). Here, we employed the HuProt arrays, each comprised of ϳ20,000 human recombinant proteins, to identify BD associated autoantibodies. Using a previously established Two-Phase strategy, we rapidly identified 20 potential BD-associated autoantigens and validated anti-CTDP1 antibody as BD-associated biomarker using a much larger cohort.
Although the pathology of BD is of great difference from the other autoimmune diseases, and let alone healthy persons, the number of autoantibody hits identified in each category is not significantly different from each other. However, the identity of the identified autoantigens varied dramatically from BD patients to the other groups. Thus, serum profiling assays performed on the HuProt arrays provide a proteome-wide fingerprints for each sample, a resolution that surpasses the traditional cell-based staining methods. Among the 20 candidate autoantigens of BD identified in the screening phase with HuProt arrays, some of which are play important role in transcription catalyzed by RNA polymerase I, such as NOLC1 and PAPD5; some of which are positive regulation of cell proliferation, such as PRAMEF12, CLPTM1, HCLS1, BHLHE22 and SH3BP1. It is very interesting SH3BP1 is act as a modulator of glutaredoxin biological activity to regulate actin cytoskeleton organization and blood vessel endothelial cell migration (31). The anti-SH3BP1 antibody in BD may affect the activity of SH3BP1 to have an important influence of actin cytoskeleton organization and blood vessel endothelial cell migration, which play important role in the pathogenesis of BD. It is also interesting to note that 11 of the 20 candidate autoantigens on BD focused arrays showed a similar sensitivity in Phase II between the BD and other autoimmune disease control cohorts, but a higher sensitivity than normal controls, suggesting that these 11 antibodies might play a similar role in the development of different autoimmune diseases. Function of these autoantibodies would be worthy further analysis. CTDP1, also known as Fcp1, is a TFIIF-associating phosphatase, which encoding gene locates in the 14 exon of the 18q23. It is ubiquitously expressed in tissues (3) and locates in nucleus and cytoskeleton in relation to cell mitosis (32). CTDP1 is known to dephosphorylate the C-terminal domain (CTD) of RNA polymerase II subunit A and promote gene expression cycle. RNAP II plays key role in messenger RNA production, during which CTD undergoes a cycle of phosphorylation and dephosphorylation. Before the preinitiation complex formation and transcription, CTD is dephosphorylated; during transcription elongation it is phosphorylated (33). CTDP1, classic phosphatase for heptapeptide repeat of CTD (34 -36), is necessary for the initiation of another mRNA synthesis cycle. CTDP1 also regulates other substrates. It inactivates crucial mitotic substrates (e.g. USP44, CDC20 and WEE1) to dephosphorylate M-phase-promoting factor (MPF)/ CDK1, which results in promoting mitosis exit (32). In addition, CTDP1 regulates transcription elongation and stimulates the rate of elongation by RNAP II (37). Taken together, anti-CTDP1 autoantibodies may neutralize the CTDP1 role of dephosphorylation including RNAP resumption and mitosis exit, and transcription enlongation, inhibiting the production of mRNA and daughter cells. This may influence tissues which metabolism quickly, such as skin and mucosa, partly explaining the recurrent ulcer of BD.
CTDP1 partial deficiency has been reported in congenital cataracts facial dysmorphism neuropathy syndrome (CCFDN), which is an autosomal recessive developmental disease (38). As the name suggested, a key factor is nervous system involvement manifested with motor neuropathy leading to disability. Moderate nonprogressive cognitive deficit and pyramidal signs are also associated neurological features. Nerve biopsy shows demyelinating pathogenesis (39). Interestingly, Neuro-Behcet Disease, accounted for 5-10% of BD patients, can also involve the nervous system, primary characterized as subacute brainstem syndrome and hemiparesis. However, unlike CCFDN mainly injure periphery system, neuro-Behcet Disease mainly damages parenchymal of the central nervous system. The major histopathologic phenomenon are vasculitic manifestation and low-grade chronic nonspecific inflammation (40). The role of CTDP1in Behcet Disease, especially in neuro-Behcet Disease, should be further investigated.
Autoimmune diseases, including Behcet's disease, SS, AAV, and TA, are highly heterogeneous. In order to save research costs, we adapted two-phase strategy combines a proteome-wide screen for novel autoantigens followed by a stringent validation step using additional large cohorts to ensure the success of identification of useful autoantigens for a particular disease. In the screening phase, only a very small number of samples (i.e. 5TA,5AAV, and 5SS) were used as the disease control groups and therefore, the sensitivity and specificity of a candidate biomarker may be different from the real situation, presumably due to over fitting. As such, the additional large cohorts(including 130 BD, 118 autoimmune disease controls, and 110 healthy controls) used for the Phase II validation is extremely important to ensure the success of identification of true autoantigens for a particular disease. Therefore, it is understandable that the final biomarker of anti-CTDP1 antibody is not on the top list in the screening phase.
In previous studies, potential candidate biomarkers, such as anti-alpha-tropomyosin, anti-S-antigen, anti-alpha-enolase and anti-selenium binding protein, were reported for BD diagnosis. Although these proteins showed decent anti-GST signals (a proxy of purified protein amount) on the HuProt arrays, we did not observe any significant signal intensity in most of the serum profiling assays with BD-positive sera. One plausible explanation of this discrepancy is that these autoantigens can only be recognized by BD-positive sera in their denatured forms. Because most of the proteins on the HuProt arrays are in their native conformation, they failed to be recognized by the BD patient sera.
In summary, our study demonstrated that anti-CTDP1 is a BD-associated autoantibody, which might be very useful in clinical diagnosis of BD in the future.